Zombie ant biting behavior likely involves an array of processes as suggested by gene expression studies.

The full research paper on this topic has been published in BMC Genomics.

Many parasites have evolved the ability to change host behavior in order to promote completion of the parasite’s life cycle. Though this phenomenon is generally assumed to be the result of gene expression, only a few studies have demonstrated its genetic basis. In our study we have explored the genetic basis of a host-parasite system in which complex parasitic manipulations are observed, that eventually lead to a wholly novel behavior being expressed by the host. We have set out to investigate which genes are possibly involved in zombie ant behavior caused by the parasitic fungus Ophiocordyceps unilateralis. In this system, infected Carpenter ants show a couple of stereotyped behaviors that ultimately lead to a manipulated biting behavior that fixes the ant to the vegetation before the fungus kills it. This biting behavior is not part of the normal behavioral repertoire and aids the fungal parasite in spreading its spores to infect new individuals.

 

Zombie ant life cycle schematic (by: Roel Fleuren, Science Transmitter)

To be able to investigate the genes that could be involved in establishing manipulated biting behavior the genomes of both the fungal parasite and the ant host come in handy. The genome (aka DNA) of a related species to our ant host had already been sequenced, so we decided to use it. However, we were not so lucky that we had similar information available about our parasite. We therefore first had to sequence the genome of our fungus, which involves the reading of the organism’s genetic information by producing many short strings. It’s kind of reading an entire book randomly, a couple of words at a time and doing this as many times as necessary to know for sure that you have read the entire book. Subsequently, all these short strings (called reads) had to be assembled in the correct order so we could make sense of it again (aka assembling the genome). Finally, a process called annotation had to be performed to determine where protein-encoding genes start and stop and what they likely code for. This led to the annotation of 7,831 protein-coding genes, which might seem like a lot, but is a bit small compared to some other fungi and especially to higher organisms (the human genome for instance contains about 20,000-25,000 protein-coding genes).

To analyze this genome, one can for instance compare it to the genomes of other organisms to study how related they are through the construction of phylogenetic trees (see Figure 1). If you look at this figure Ophiocordyceps unilateralis is most related to Ophiocordyceps sinensis, which infects caterpillars but does not induce a dramatic behavioral manipulation as seen in the ants.

 

Figure 1 Phylogenetic tree comparing manipulating parasite Ophiocordyceps unilateralis with related non-manipulating fungal species (Source: BMC Genomics)

Now, regarding behavioral manipulation, we expect the fungus to release proteins and metabolites that affect the host’s brain. So, we expect it to contain many genes that encode for proteins that are being secreted (in fact, this is the case for fungi in general). This signal for secretion is part of the genetic coding for a protein so we can specifically search for these genes in the genome. When we did this, we found that our manipulating fungus has indeed many genes encoding for secreted proteins and that over 30% of these genes are unique to this fungus when compared to the fungal species mentioned in Figure 1. We also found that a large part of this group are SSPs (small secreted proteins). These type of proteins are generally found to be interesting bioactive compounds so this is a group we will definitely look closer into in the future to figure out what their functions are.

Another striking thing we found were the high amount of genes encoding for secreted enterotoxins. These toxins are generally found in bacteria and have only been reported for some fungi. However, the amount of genes (34 in total!) in our parasitic manipulator’s genome was much higher compared to those! The literature reports that parasitic enterotoxins might be impairing chemosignalling in insects, which is a very important part of ant communication. This might thus explain the “anti-social” behavior we tend to observe in our infected ants.

With the fungal genome now also in our toolbox we then set out to examine how these genes are being put to use during manipulated biting behavior. We therefore isolated RNA from heads of zombie ants while they were biting, and after they had died. The RNA gives us information about how genes are put to use in a certain situation i.e. de expression level of the genes. This RNA is subsequently translated into proteins, which can be involved in all sorts of processes. The gene expression levels thus give us an indication of the processes that are employed during a certain situation, such as biting behavior. We therefore need to be able to compare it to gene expression levels of so-called controls to determine if certain processes are activated or deactivated during biting behavior. As controls, here, we isolated RNA from healthy ants and the fungus grown outside of the ant host (Figure 2).

 

Figure 2 Summary of sample types used in our study

As a next step we also sequenced the RNA, kind of in the same way as we did for the genome by generating a lot of “reads”. These reads were subsequently mapped to the genome of the ant and/or the fungus to figure out how much RNA was present for each gene in each of the samples. To try to visualize this for you, have a look at Figure 3. After this work was all done, we performed some statistics to determine which genes and processes significantly changed in their activity when comparing the different samples / situations to each other, i.e., we determined the differential expression patterns.

 

Figure 3 Schematic explaining the mapping of RNA reads to the fungal parasite and ant host genome prior to determining the changes in gene expression between samples

Looking at the differential expression patterns, we found that 498 of the fungal genes that were significantly higher expressed during manipulated biting behavior were down-regulated again after the manipulated ant had died. These genes might thus encode for proteins that are specifically used by the fungus to induce manipulated biting. Interestingly, when searching if these genes also exist within the genomes of fungal insect parasites that do not manipulate behavior, we found that 80% (so 410 of these genes) appear to only exist in the genome of Ophiocordyceps unilateralis. These fungi might thus indeed have evolved some very unique abilities to deal with their hosts!

Of course we also had a closer look at the possible functions of the proteins that these genes encode for. Moreover, we also investigated the gene expression taking place in the ant host. Below you can find a figure summarizing our findings, but if you’re interested in the individual genes and their expression patterns, have a look at Table 2 and 3 of our paper.

 

Summary of candidate proteins and pathways found in this study that could possibly be involved in establishing manipulated biting behavior as seen in zombie ants (by Roel Fleuren, Science Transmitter)

As discussed above, our study also showed that O. unilateralis s.l. has an abundance of genes encoding for bacterial-like enterotoxins, which are dynamically up- and down-regulated in the situations tested in this study. These compounds could negatively affect the production of chemo signaling molecules in insects, which would impair communication in infected individuals. In agreement with this, infected ants are generally observed to be unresponsive to external stimuli and host genes encoding for receptors for these compounds were found to be down-regulated. However, the dynamic expression of these O. unilateralis s.l. enterotoxins could also be contributing to the signs of programmed cell death found in manipulated C. castaneus ants and the severe atrophy of muscle cells observed in earlier studies.

Genes involved in host immune responses were found to be generally down-regulated in infected ant heads during the manipulated biting event. We hypothesized that this suppression of host responses might be a way for the fungal parasite to create conditions in the host that favor parasite development. While O. unilateralis s.l. ultimately kills its host, premature death will not lead to manipulation thus badly affect the chances for the parasite to spread its spores and complete its life cycle. In addition to causing apoptosis and impairing chemosensory communication, O. unilateralis s.l. might therefore be regulating the host’s immune- and neuronal stress responses as a mechanism to establish manipulated behavior.

Other links found between the fungal parasite’s and ant host’s gene expression during manipulated biting behavior are the up-regulation of genes encoding for enzymes involved in dopamine metabolism and protein-tyrosine phosphatases. Both pathways have reported effects on behavioral outputs such as locomotion and mandible movement, which are hallmarks of ant infections by O. unilateralis s.l. fungi. Moreover, the expression of certain clock genes (these make sure our biological rhythms are in sync with the daily changes in light and temperature) was changed in manipulated ants. This would affect processes in the host, ranging from the cellular to the behavioral output level.

In addition, we found the up-regulation of various fungal genes involved in different types of alkaloid metabolism (for example, morphine, nicotine and LSD are different types of alkaloids) that lead to compounds that can bind to receptors in the central and peripheral nervous system and this way block or activate neuronal pathways. We also found a higher expressed secreted sphingomyelinase that changes neuron cell membranes, a possible shift in tryptophan metabolism, which could lead to behavioral changes and the production of an array of bioactive metabolites and small secreted proteins with unknown functions.

We thus found various mechanisms and compounds through which the ant host brain could be manipulated and manipulated biting behavior in zombie ants could be established. While some of these candidates could be essential to the process, we expect these to work together since the manipulated behavior in this system is so complex and precise. Future functional gene expression studies are however needed to determine the importance and the exact function of the genes found in this study. The work presented here thus forms an important stepping stone towards truly unraveling how a parasite can create zombie-like behavior. It also paves the beginning of the road towards a better understanding of how “normal” behavior is regulated, and what might tip the scale towards “pathological” behaviors. Moreover, the impressive track record of fungi producing industrially interesting bioactive compounds and the expectation that brain-manipulating species produce novel neuromodulators, also make genetic studies into this system important from an applied perspective.