Zombie ant death grip due to hypercontracted mandibular muscles

Parasites are capable of expressing their phenotypes in the bodies of the hosts that they occupy. This is an example of an extended phenotype (Dawkins, 1982). Specific viruses within the Baculoviridae family including Lymantria dispar nucleopolyhedrovirus (Hoover et al., 2011) and Mamestra brassicae nucleopolyhedrovirus (Goulson, 1997) infect L. dispar and M. brassicae larvae, respectively. This triggers distinct changes in behavior where infected larvae climb up to higher elevations on vegetation, die and liquify. This process is hypothesized to positively impact the transmission of infectious progeny virions to other hosts (Goulson, 1997; Heil, 2016). Precise manipulation of host behavior is achieved via the coordinated expression of specific viral genes including ecdysteroid UDP-glucosyltransferase, protein tyrosine phosphatase, chitinase and cathepsin (Han et al., 2015; Hawtin et al., 1997; Hoover et al., 2011; Kamita et al., 2005; Katsuma et al., 2012). Another example is the lancet liver fluke Dicrocoelium dendriticum, which infects a number of different grazing animals, but requires transmission from mollusks to intermediate ant hosts for maturation prior to transmission to their ruminant definitive hosts (Manga-González et al., 2001). Once infected, ants display a change in behavior where they climb to the tops of blades of grass and bite down during the early dawn and evening hours (Botnevik et al., 2016; Manga-González et al., 2001). It has been suggested that D. dendriticum expresses this extended phenotype via a direct influence on the brains of their ant hosts (Botnevik et al., 2016; Carney, 1969; Krull and Mapes, 1952; Martín-Vega et al., 2018; Moore, 1995).

One of the most well-studied microbial manipulators in insects is the entomopathogenic fungus Ophiocordyceps unilateralis sensu lato. Infection with O. unilateralis s.l. triggers a pronounced change in behavior where infected ants climb up vegetation and bite onto leaf veins or twigs, where they ultimately die (Andersen et al., 2009; Hughes et al., 2011; Loreto et al., 2018). The fungus grows within the ant's body, consuming all of its resources, extends a stalk from the ant's head and then sporulates, dispersing infectious spores from the stalk onto uninfected ants foraging along the forest floor below. The fungus is unable to grow or spread within the nest, making the manifestation of these behavioral changes in areas around individual colonies essential for efficient transmission to new hosts (Loreto et al., 2014). The ‘zombie-ant’ phenotype is complex and is composed of two distinct behaviors, (1) climbing up vegetation and (2) the ‘death grip’, where the ant bites onto a leaf vein or twig prior to death. Previous studies examining infected host mandibular muscle morphology show extensive muscle cell atrophy and colonization of the head capsule by fungi at the moment of manipulation (Fredericksen et al., 2017; Hughes et al., 2011). Moreover, within the host, cells of Ophiocordyceps kimflemingiae (=unilateralis; Araújo et al., 2018) act cooperatively and form an interconnected community around individual mandibular muscle cells in the infected ant. This causes the muscle fibers to become widely separated as the invading organism occupies up to roughly 40% of the biomass of the ant host (Fredericksen et al., 2017).

Vertebrate muscle cells propagate signals from neuromuscular junctions (NMJs) across the length of the muscle, resulting in full muscle contraction. In contrast, invertebrate muscle cells can be innervated by multiple motor neurons at a number of different points along the length of the muscle cell (Edwards, 1959; Usherwood, 1967). Contraction of ant mandibular closer muscle is controlled by 10–12 motor neurons (Paul and Gronenberg, 2002). In the cockroach abdominal muscle, NMJs are spaced 4–30 µm apart, with each muscle fiber possessing 1–4 junctions (Edwards, 1959). Although not definitively quantified in ant mandibles, this suggests that ant mandibular muscle should have numerous contacts with motor neurons.

Prior observations have demonstrated extensive atrophy and muscle invasion at the time of the death grip behavior in O. kimflemingiae-infected ants (Fredericksen et al., 2017; Hughes et al., 2011). These observations combined with our understanding of the architecture of neuromuscular systems in insects raise different questions. For example, how can such precise manipulation of host behavior occur when the muscle is extensively atrophied? And how can motor neurons and NMJs be maintained when the muscles are so heavily invaded by the fungal parasite? Answering these questions may provide important insight concerning the mechanisms utilized by O. unilateralis s.l. to trigger the death grip. Given the dense, interconnected networks formed by O. kimflemingiae around muscle cells (Fredericksen et al., 2017) and the previously documented muscle cell atrophy (Hughes et al., 2011), we hypothesized that motor neurons innervating the mandibular muscle may become degraded as the fungal cells begin to crowd the intermuscular space. The fungus could then act locally to trigger muscle contraction.

To investigate overall muscle morphology and test for the presence and integrity of motor neurons, we analyzed mandibular muscle from O. kimflemingiae-infected ants using scanning electron microscopy (SEM) to visualize gross muscle, motor neuron and fungus structure. We then utilized transmission electron microscopy (TEM) to assess the effects of infection on muscle cytoarchitecture and test for the presence of nervous tissue. As controls, we used healthy ants and ants infected with the fungus Beauvaria bassiana, a generalist and non-manipulative entomopathogenic fungus within the same order (Hypocreales) as O. kimflemingiae. Our data indicate that at the moment of biting, mandibular muscles of zombie ants demonstrate hypercontraction. Despite the abundance of fungus surrounding muscle cells, motor neuron innervation was maintained and NMJs were present. Sarcolemma breakdown was evident; however, this also occurred in the mandibular muscle of B. bassiana-infected ants. Further investigation identified potential alternative mechanisms underlying muscle hypercontraction at the time of biting. These include the formation of interconnected fungal cell networks, direct insertion of hyphal tubes into muscle and the secretion of extracellular vesicle-like particles. Together, the data presented here suggest that the zombie ant death grip phenotype may be a result of O. kimflemingiae-specific secretion or direct introduction of modulators that mediate muscle contraction. Compounded with extensive damage to the sarcolemma, this may cause muscle hypercontraction. These results provide an important step towards understanding the mechanisms underlying the manifestation of the zombie ant death grip.

Camponotus castaneus (Latreille 1802) ants were collected in Donalds, SC, USA, and housed in plastic containers (Sistema, Mangere Auckland, New Zealand) within a temperature-controlled room (25.8–23.2°C). The insectary was maintained on a 12 h:12 h light:dark cycle with a relative humidity of roughly 79–99%. Water and 10% sugar water were available ad libitum and replenished every 1–2 weeks. Biting platforms made of Plaster of Paris (DAP, Baltimore, MD, USA) and embedded with autoclaved toothpicks (Diamond, Jarden Home Brands, Fishers, IN, USA) were placed inside each cage to provide a substrate for climbing and biting following behavioral manipulation. Nests were made using a Plaster of Paris base with a red transcolor PVC film (TAP Plastics Inc., Sacramento, CA, USA) over an empty pipette tip box lid (200 µl, VWR Zap, Radnor, PA, USA). Ants were randomly assigned to a treatment group (infected or control) and were individually painted with a paint marker (Edding 751, Ahrensburg, Germany) to distinguish group assignment.

Manipulated ants were collected from two separate infections. For the first infection, O. kimflemingiae was harvested from the gaster (part of the thorax) of a recently deceased C. castaneus ant. Fungal tissue was incubated on a potato dextrose agar media plate for 10 months until growth reached roughly 1.5 cm in diameter. Inoculum was prepared as previously described (de Bekker et al., 2014). Roughly 0.375 cm² of fungal tissue was excised, homogenized in 500 µl Grace's insect medium (Sigma-Aldrich, St Louis, MO, USA) and diluted 4-fold. Ants were injected with 1 µl of inoculum into the membrane located underneath the forelegs using a laser-pulled 10 µl micropipette (Drummond Scientific Co., Broomall, PA, USA) and aspirator tube (Drummond Scientific Co.).

The second infection was executed as follows. Deceased ants biting onto twigs were collected in Donalds, SC, USA. Following surface sterilization with 70% ethanol, the cuticle was removed and all internal tissue containing O. kimflemingiae was placed into a sterile conical 1.5 ml tube (VWR) with two sterile 5/32 inch metal balls (Wheels Manufacturing Inc., Louisville, CO, USA) and 100 µl Grace's Insect Media (Thermo Fisher Scientific, Hampton, NH, USA) supplemented with 10% fetal bovine serum (Life Technologies, Carlsbad, CA, USA). Fungal material was then lysed for 1 min at 30 Hz using a Tissue Lyser II (Qiagen, Germantown, MD, USA), and briefly centrifuged using a tabletop microcentrifuge (Benchmark Scientific, Sayreville, NJ, USA) to remove cellular debris. The resulting supernatant was used as the inoculum; 1 μl of inoculum was then injected into C. castaneus ants as described above. Control ants (n=6) were injected with 1 µl Grace's Insect Media supplemented with 10% fetal bovine serum. Infection was monitored for roughly 30 days with behavioral manipulation occurring most often between days 15 and 30. Ants were identified as behaviorally manipulated if they were biting onto a substrate, exhibited muscle spasms, did not respond to external stimuli and did not detach from the substrate when touched (TEM, n=3; SEM, n=6). At the time of biting when the ants were still alive, they were immediately flash frozen or dissected for histology.

Positive control ants were infected with B. bassiana, an entomopathogenic fungus that does not induce the death grip behavior. Beauvaria bassiana conidia were provided by Dr Nina Jenkins at Penn State University. Infectious conidia were suspended in 0.05% Tween/phosphate-buffered saline, poured onto filter paper, and placed in a 60 mm Petri dish. Camponotus castaneus ants were then placed in the Petri dish and allowed to walk over infectious spores for 24 h. Infected ants were then placed in a new cage. For B. bassiana-infected ants, there are no stereotypical behavioral phenotypes like those observed in O. kimflemingiae-infected ants that happen prior to host death, which occurs between 3 and 5 days post-infection. Infected ants may exhibit sickness behaviors well before death, making it difficult to correctly identify ants that will die within the next few hours. As such, between days 3 and 5 post-infection, infected ants were periodically monitored in 1 h intervals to ensure collection of ants within 1 h of death. Ants were then flash frozen in liquid nitrogen until processing for histology (control, n=3; infected, n=3).

For samples analyzed by SEM (n=3–6 per group), mandibular muscle still attached to cuticle was rapidly dissected and placed in 2.5% glutaraldehyde in 0.1 mol l⁻¹ sodium phosphate buffer overnight at 4°C. Tissue was then washed 3 times with 0.1 mol l⁻¹ sodium phosphate buffer for 5 min each, dehydrated with 25%, 50%, 70%, 85% and 95% ethanol once each for 5 min, and then 100% ethanol 3 times. Dehydrated samples were then dried by critical point drying (Leica EM CPD300, Wetzlar, Germany). Images were acquired using a Zeiss Sigma VP-FESEM (Thornwood, NY, USA). To assess average extracellular vesicle diameter, 100 particles per group were measured using Fiji (ImageJ).

For samples analyzed by TEM (n=3 per group), mandibular muscle was immediately removed, placed in modified Karnovsky's fixative (2.5% paraformaldehyde, 1.5% glutaraldehyde, 0.1 mol l⁻¹ sodium cacodylate buffer) and incubated overnight at 4°C. Tissue was then washed 3 times in 0.1 mol l⁻¹ sodium cacodylate buffer for 5 min and fixed in 2% osmium tetroxide for 1 h protected from light. Samples were washed twice in 0.1 mol l⁻¹ sodium cacodylate buffer, once in Milli-Q water, and then stained for 1 h in 2% aqueous uranyl acetate protected from light. Samples were dehydrated in ethanol (50%, 70%, 85%, 90%, 95% and 100% once each for 5 min, and 3 times in 100% EM grade ethanol), washed 3 times in acetone, and then infiltrated with Spurrs as follows: 50% acetone/50% Spurrs overnight, 25% acetone/75% Spurrs ∼8 h, 100% Spurrs overnight, 100% Spurrs ∼8 h, 100% Spurrs overnight, and then polymerized in fresh Spurrs overnight at 60°C. Ultrathin sections (70 nm, Leica EM UC6 Ultramicrotome) were placed on grids, post-stained in 2% uranyl acetate/50% ethanol, washed in distilled water, and then stained in lead citrate. Grids were washed, allowed to dry, and then imaged on a JEOL JEM 1200 EXII (Peabody, MA, USA).

Previous research has shown the dense network of fungal cells infiltrating mandibular muscle in infected ants at the time of biting (Fredericksen et al., 2017). In support of these previous data, SEM analysis of uninfected (Fig. 1A) and infected (Fig. 1B,C) mandibular muscle demonstrated the diffuse presence of fungal cells on and around individual muscle fibers of manipulated ants. This phenotype was also seen in B. bassiana-infected ants, where fungus was observed throughout the mandibular muscle space (Fig. S1). Numerous muscle cells in O. kimflemingiae-infected ants demonstrated a unique morphology where z-lines appeared to be swollen and the sarcomeres shortened, giving a grooved-like appearance (Fig. 1B,C). Additionally, extensive damage to the sarcolemma was evident (Fig. 1C, arrowheads). Closer inspection of these regions of sarcolemma breakdown revealed what seem to be individual myofibrils beneath the membrane (Fig. 1D, arrow). Areas presumed to be the z-lines (Fig. 1D, arrowheads) demonstrated significant damage, wherein the fibers appeared frayed and broken.

Given the distinct change in muscle cell appearance at the time of the death grip, we wanted to investigate the structural effects of O. kimflemingiae infection on muscle cells, using TEM to visualize intracellular muscle fiber ultrastructure. As muscle from B. bassiana-infected ants more closely resembled control muscle, where z-line shortening was not as prevalent nor as severe, we did not pursue further analysis of these samples by TEM. In O. kimflemingiae-infected muscle, we were able to observe distinct changes in z-line morphology at the time of biting. The z-lines in control ants exhibited a connected, regular, linear morphology (Fig. 2A). In contrast, infected muscle demonstrated changes in z-line morphology where they were swollen (Fig. 2B), irregular and/or disengaged from neighboring z-lines (Fig. 2C), suggesting sarcomere disruption and shortening. Additionally, the presence of numerous myelin-like structures was evident within infected muscle (Fig. 2B, arrows).

We next sought to examine the presence and integrity of motor neurons and NMJs within the mandibular muscle. It was more difficult to locate motor neurons and NMJs in infected tissue than in control tissue (Fig. 3A), but we were able to identify these structures and see that they remained intact (Fig. 3B). This was also true in B. bassiana-infected muscle (Fig. S2). In some cases, O. kimflemingiae cell clusters were observed in direct contact with motor neurons and near NMJs (Fig. 3B). Unfortunately, surface imaging did not allow for determination of the presence of motor neurons within fungal cell clusters. We were therefore unable to quantify exact differences in motor neurons between infected mandibular muscle and healthy controls. Nonetheless, in both control and infected ants (Fig. 3C,D, respectively), nervous tissue was present around and in contact with muscle cells.

One unanticipated finding was that in the O. kimflemingiae-infected mandibular muscle samples examined, we observed gross degradation of muscle fiber sarcolemma (Figs 1C,D and 4A). In most cases, these sections of sarcolemma breakdown were located directly beside clusters of fungal cells. Similar to O. kimflemingiae infection, B. bassiana infection also triggered sarcolemma degradation (Fig. 4B). Some small patches of sarcolemma destruction were observed in control tissue. However, these sections were most often localized near excision points of the dissection and are more than likely an artifact of the dissection itself. Degraded sarcolemma was much more prevalent in infected muscle and was localized to internal muscle fibers as well, suggesting these observations are not artifacts.

Using SEM, we were also able to observe the previously documented collective behavior of fungal cells: anastomosis tubes connecting individuals, resulting in the formation of fungal networks around host muscle cells (Fredericksen et al., 2017) (Fig. 5, arrows). Individual fungal cells can form multiple connections along the length of the entire cell body (Fig. 5A). Additionally, we were able to observe the formation of these connection points (Fig. 5A, arrowheads).

Consistent with previous literature (Fredericksen et al., 2017), we were able to observe numerous hyphal bodies in direct contact with ant muscle (Fig. S3), as well as hyphal body invasion of muscle cells (Fig. 6). In these cases, invading hyphal bodies were connected to others via extensive anastomosis bridge networks. Additionally, we identified the presence of numerous extracellular vesicle-like particles associated with fungal cells within the mandibular muscle space (Fig. 7, arrows). Extracellular vesicle-like particles were also observed in B. bassiana-infected muscle (Fig. S4). Extracellular vesicle-like particles identified within O. kimflemingiae-infected muscle were small (averaging 463 nm in diameter), spherical, similar in size and often found in small (Fig. 7B) or large (Fig. 7D) clusters. In contrast, vesicles found within B. bassiana-infected tissue appeared larger (averaging 1.6 µm in diameter), less spherical and were randomly dispersed (Fig. S4).

Manipulation of host behavior by O. unilateralis s.l. is a complex phenomenon that is manifested by the stereotypical ‘death grip’ where infected ants bite into vegetation prior to death (Andersen et al., 2009; Hughes et al., 2011). Studying how this phenotype arises is of significant interest and will yield important insight into pathogen–host co-evolution as well as potential mechanisms underlying fungal pathogenesis. In this study, we sought to understand the interactions between host muscle tissue and fungal cells.

Prior studies examining the interactions between entomopathogenic fungi and their hosts have demonstrated that in general, despite the widespread distribution of fungal cells around host muscle, destruction and physical invasion are delayed until after death (Elya et al., 2018; Funk et al., 1993; Gryganskyi et al., 2017). This potentially allows the infected hosts to continue moving around and consuming nutrients until the time of death (Gryganskyi et al., 2017). In Ophiocordyceps-infected ants, our lab has shown the development of an extensive, interconnected, fungal community around individual muscle fibers in host mandibles, suggesting that fungal cells may work cooperatively to consume host nutrients and induce the biting behavior seen in infected ants (Fredericksen et al., 2017). In the present study, we built on these previous findings to provide a more global view of fungal–host interactions in this model system (see Table 1 for a list of similarities and differences between the two studies and fungal pathogens used). We confirmed that host mandibular muscle is heavily populated by O. kimflemingiae cells (Fig. 1B). In addition, we observed numerous structural abnormalities in infected host cells, the most striking of which was z-line swelling and sarcomere shortening (Fig. 1B–D).

In order to investigate this phenotype further, we utilized TEM to characterize changes to muscle cell ultrastructure at the time of biting in O. kimflemingiae-infected ants. Areas of z-line swelling, shortening (Fig. 2B) and disruption (Fig. 2C) were evident in infected muscle. Swelling and disruption of the z-line as well as sarcolemmal corrugation have been observed in human skeletal muscle following eccentric exercise, where the muscle concomitantly lengthens as it contracts (Lauritzen et al., 2009), and is indicative of muscle hypercontraction. Similar phenotypes are also observed in insect striated muscle following intensive contraction. In adult female locusts, Locusta migratoria, sarcolemma corrugation and z-line disruption are evident in supercontracted ovipositor muscle in preoviposition females (Jorgensen and Rice, 1983). Additionally, corrugation of the sarcolemma and the formation of myelin-like bodies (Fig. 2B) are abundant in O. kimflemingiae-infected muscle. The appearance of myelin-like structures occurs in rat heart muscle following administration of adrenalin (Kleimenova and Belen'kii, 1975), suggesting that these bodies form in response to hypercontraction. Together, the data presented in the present study detailing z-line swelling, sarcomere shortening and disruption, sarcolemma blebbing and the formation of myelin-like bodies in infected muscle (Fig. 2) support the hypothesis that mandibular muscle at the time of biting is in a state of hypercontraction. This phenomenon had not been previously reported in other studies examining entomopathogenic fungus–insect interactions.

The motor neurons and NMJs of manipulated ants appear to be present and structurally intact (Fig. 3). This is true despite extensive colonization of fungal cells around individual muscle fibers and the wide separation of muscle cells evident at the time of biting. Although we observed fewer motor neurons and NMJs, this may be due to the large number of fungal cells present in the mandibular muscle space. Alternatively, motor neurons may degenerate over the course of infection, yielding a smaller number at the time of biting. The host nervous system is suggested to be important for behavioral manipulation (Adamo, 2013). In fact, the central nervous system of the manipulated ants appears to be physically intact at the moment of the death grip (Fredericksen et al., 2017), and it is arguable that motor neurons are also preserved at this time. Therefore, our difficulty in finding motor neurons and NMJs may be a consequence of clusters of fungal cells that were physically obstructing our view (Fig. 1B).

When assessing the morphology of infected mandibular muscle, we observed overt damage to the sarcolemma localized throughout the mandibular muscle space in all samples infected with O. kimflemingiae (Fig. 4A). We hypothesized that perhaps this diffuse degradation of the sarcolemma may underlie aberrant muscle contraction observed in manipulated ants. However, we observed similar regions of sarcolemma degradation in the mandibular muscles of B. bassiana-infected ants (Fig. 4B). Therefore, muscle membrane damage is not necessarily specific to infection with O. kimflemingiae or the sole mechanism underlying initiation of the death grip. Nonetheless, sarcolemma degradation may still play a role in initiating the death grip associated with manipulation.

The sarcolemma is responsible for maintaining proper ionic balance and resting membrane potential (Hopkins, 2006). Once stimulated by excitatory motor neurons, the membrane potential depolarizes, releasing calcium stores from the sarcoplasmic reticulum, which triggers muscle contraction (Iwamoto, 2011; Josephson, 2009). Disruption of the sarcolemma can cause aberrant calcium influx, resulting in alterations in muscle contraction and ultimately muscle fiber degradation, as occurs in Duchenne muscular dystrophy (Petrof, 2002). Degradation of the sarcolemma is also thought to contribute to muscle hypercontraction in response to notexin, a toxin in Australian tiger snake (Notechis scutatus scutatus) venom (Dixon and Harris, 1996). It is unlikely that the sarcolemma damage directly causes the hypercontraction of mandibular muscles observed during the death grip because it is also abundant on non-manipulated ants infected with B. bassiana. However, sarcolemma breakdown could facilitate the direct introduction of fungal secreted factors to trigger the observed hypercontraction.

At the time of biting in O. kimflemingiae-infected ants, fungal genes associated with ergot alkaloid biosynthesis are increased in expression, suggesting these compounds may play a role in behavioral manipulation (de Bekker et al., 2015). Claviceps purpurea, a fungus closely related to Ophiocordyceps, secretes ergot alkaloids, compounds known to trigger alterations in serotonergic neurotransmission, resulting in convulsions and hallucinations in mammals (Eadie, 2003). Additionally, the entomopathogenic fungus Metarhizium anisopliae releases destruxins when infecting tobacco hornworm (Manduca sexta) larvae. These destruxins cause rapid depolarization of host muscle membrane, potentially via modulation of calcium channels, thereby triggering sustained muscle contraction and paralysis, followed by flaccid paralysis (Samuels et al., 1988). The O. kimflemingiae genome encodes 36 putative enterotoxins (de Bekker et al., 2017). Importantly, transcription of one putative enterotoxin ortholog is up-regulated ∼3000-fold at the moment of biting, with transcription decreasing ∼200-fold following the death grip (de Bekker et al., 2015, 2017). This gene is encoded in numerous Ophiocordyceps species that are able to manipulate ant behavior (de Bekker et al., 2017). These data suggest that enterotoxins may play important roles in the manifestation of the death grip. However, these toxins still need to be isolated from O. kimflemingiae and fully characterized. Further study of these fungus-derived toxins is required to understand their potential involvement in disease pathogenesis and manifestation of the death grip.

We observed tight association of fungal cells to host sarcolemma (Fig. S3), invasion of host muscle by fungal hyphae (Fig. 6), and the presence of fungus-associated extracellular vesicle-like particles (Fig. 7). Physical invasion of host muscle by O. kimflemingiae may provide a mechanism by which the fungus can extract nutrients from host tissue and feed the network of fungal cells via anastomosis tube connections. This is supported by data demonstrating invasion of Frankliniella occidentalis thorax muscle by Verticillium lecanii post-mortem. Areas of tissue clearing are evident around hyphal tips within muscle tissue, indicating enzymatic degradation (Schreiter et al., 1994). A similar process may be utilized by O. kimflemingae, where invading hyphae may digest local muscle tissue, and nutrients can then be shuttled through the interconnected network of fungal cells.

Further research is required to determine whether the identified extracellular vesicle-like particles are derived from host cells in response to infection, or are fungus derived. At least eight species of fungi have been shown to release extracellular vesicles, including the human pathogenic fungi Cryptococcus neoformans, Histoplasma capsulatum, Paracoccidiodes brasiliensis and Candida albicans (Albuquerque et al., 2008; Gehrmann et al., 2011; Oliveira et al., 2013; Rodrigues et al., 2007, 2011; Vallejo et al., 2011; Vargas et al., 2015; Zamith-Miranda et al., 2018). The release of extracellular vesicles may play important roles in modulating the interactions between fungi and their hosts. Specifically, they may contribute to fungal pathogenesis and virulence (Huang et al., 2012; Oliveira et al., 2010, 2013; Panepinto et al., 2009; Rodrigues et al., 2008, 2011; Zamith-Miranda et al., 2018). These findings combined with the data presented here suggest that perhaps O. kimflemingiae secretes putative enterotoxins packaged within extracellular vesicles in a synchronous fashion that triggers local mandibular muscle hypercontraction. We observed similar invasion by B. bassiana as well as evidence of extracellular vesicle-like particles in B. bassiana-infected mandibular muscle (Fig. S4). Further testing needs to be done to determine whether B. bassiana- and O. kimflemingiae-derived extracellular vesicles contain similar or unique cargo. These data would provide valuable information detailing potential mechanisms of distinct pathogenesis strategies.

Unraveling the mechanisms by which a microbe is able to manipulate host behavior will aid in elucidating the fundamental mechanisms underlying animal behavior as well as detail mechanisms of fungal pathogenesis and virulence. In the case of zombie ants, many questions remain unanswered. In the present study, we sought to investigate how despite the extensive fungal colonization and atrophy of mandibular muscles, manifestation of the death grip still occurs. We were able to show that rather than exhibiting overt atrophy, muscle cells are in a state of hypercontraction and motor neuron innervation is maintained. Additionally, individual O. kimflemingiae cells form a collective network around muscle cells, can physically invade the muscle and possess numerous extracellular vesicle-like particles on their surface. Taken together, we hypothesize that potential mechanisms utilized by O. kimflemingiae to trigger synchronized mandibular hypercontraction may include sarcolemma degradation and direct action on host muscle by secreted factors. The results presented here provide a strong foundation from which to begin investigating the contribution of these different proposed mechanisms to muscle hypercontraction and the manifestation of the zombie ant death grip.

We would like to thank the faculty and staff members of the Huck Microscopy Facility at Penn State University including John Cantolina and Dr Greg Ning for their support and guidance during all the imaging experiments. We would also like to thank Drs Vivian Budnik and Wulfila Gronenberg for their invaluable insight on, and assistance with data interpretation.