Host manipulation by parasites: a spooky Halloween post

Happy Halloween! Turn on some spooky music, grab a handful of candy, and settle in for a spooky post about host manipulation by parasites! But do be careful eating; you never know when a parasite might be sneaking into your diet.

Cartoon of snails that look like Halloween candy because they have been manipulated by parasites.

What is adaptive manipulation?

If you’re a parasite enthusiast, you’ve probably heard about parasites that manipulate host behavior. But what, exactly, is host manipulation? We expect that infected hosts will often act “sick”; they might try to fight off their infections by raising their body temperature (fever) or eating medicinal plants, or they might be suffering from so much pathology from their parasites or immune responses that they lose their appetites, act lethargic, etc. But none of those sickness behaviors are adaptive manipulation. To be adaptive for the parasite, the changes to the host’s phenotype (coloration, morphology, and/or behavior) must produce a fitness benefit for the parasite, usually by increasing parasite transmission from one host to the next. Following this definition, the host’s phenotype is actually the parasite’s extended phenotype: the parasite’s genes control how the host looks or behaves.

To show that a host’s phenotype is an example of adaptive manipulation, we need to show that that the phenotype actually increases parasite transmission. This can be tricky! But here’s an example. Last week, we talked about the life cycle of Euhaplorchis californiensis, a trematode with a complex life cycle. The trematode infects killifish as the second intermediate host; specifically, it infects the hosts’ BRAINS. Infected killifish are more likely to do flashy swimming behaviors, like dashing up to the surface. This seems like an adaptive manipulation that would make infected fish more likely to get eaten by their definitive bird hosts; but how would you prove that? If you’re Kevin Lafferty, you’d put a bunch of killifish in two cages in the salt marsh, leaving one open to birds and one closed to birds. You can hear Kevin describe the experiment in this TED talk. At the end of the experiment, the cage that was open to birds had fewer infected fish than the cage closed to birds, because the infected fish with the flashy behaviors were more likely to be eaten than uninfected fish. The manipulated behavior increased transmission by increasing predation.

Most parasites that manipulate host behavior are not as well studied as Euhaplorchis californiensis. Therefore, we should be careful not to overstate adaptive benefits where they do not exist or have not been demonstrated, or to overdramatize demonstrated adaptive behaviors. For example, we often say that crickets infected with nematomorphs commit suicide by jumping into water, but not all crickets die or even lose the ability to reproduce after being infected. Furthermore, infected crickets aren’t so water-fixated that they hop in a straight line to the nearest waterbody; they’re just more likely to end up in water than uninfected crickets. It’s important to remember that many details tend to be glossed over when telling exciting and spooky stories, including those about parasites that manipulate host behavior.    

What kinds of manipulation exist?

We’ve mentioned a few examples, but how common is adaptive manipulation among parasites? We don’t know exactly how many parasites have evolved to manipulate their hosts, but we do know of a few hundred examples spread across many parasite groups (platyhelminths, acanthocephalans, nematodes, nematomorphs, arthropods, bacteria, fungi, protozoa, and even viruses). In fact, host manipulation mechanisms seem to have evolved independently many times within and between parasite groups. From this, we can conclude that adaptive manipulation is common, despite seeming so biologically bizarre.

But of course, there is spectacular variation among parasites in the ways that they manipulate their hosts. We can divide these methods into roughly five manipulation targets: predator avoidance, habitat choice, feeding or foraging, bodyguard behavior, and contact rates. I’ll explain each below.

Predator avoidance. Host manipulation seems especially common for parasites with complex life cycles that involve trophic transmission, where an intermediate host is consumed by a definitive host. In these systems, parasite transmission may be increased whenever the intermediate host becomes more conspicuous or less defended from the definitive host. Euhaplorchis californiensis is a good example of a parasite the manipulates how prey behave in the presence of predators.

Habitat choice. Sometimes parasites with complex life cycles have host species that do not share the same habitats or propagules (environmental stages) that require different habitats than the hosts, making it tricky for the parasite to make it from one host to the next. For example, larval nematomorphs infect terrestrial crickets and other arthropods, adults are free swimming in aquatic environments, and then larvae need to make it back to terrestrial environments to infect more arthropods. Nematomorphs manipulate cricket behavior to make them more likely to jump into water bodies—habitats that they would not usually choose. The story for how nematomorphs make it back to terrestrial environments might be even weirder, but the details are still being studied.

Host feeding behavior. Many parasites are consumed and/or transmitted during feeding, like those that are transmitted by blood-feeding insect vectors. If a vector only ever takes one blood meal, it cannot move parasites from one host to the next. To increase transmission opportunities, several parasites are known to affect vector feeding behavior, causing vectors to feed on more hosts and/or change their feeding behavior to increase the likelihood that a given feeding event will lead to transmission. A very different example might be trematodes that castrate their first intermediate snail hosts; snails reduce the time and energy that they spend reproducing when they are castrated, and instead spend that time eating, growing larger, and making more tissues available for the parasites to turn into more parasites. You can more read about parasitic castrators and gigantism in this blog post, including whether gigantism is truly adaptive for parasites.

Bodyguard behavior. Some parasites cause their hosts to co-opt normal behaviors or use abnormal behaviors to protect their parasites. For example, parasitoid wasp larvae cause orb weaving spiders to construct a protective web, and after eating the spider, the wasp larvae use the protective web as a safe hide out while they pupate. If you don’t go Google parasite bodyguards after reading this, you aren’t doing Halloween right.

Contact rates. Finally, although parasites with complex life cycles are usually the focus of adaptative manipulation research, parasites with direct life cycles and direct transmission might also benefit from manipulating host behavior. For example, if sexually transmitted parasites can cause their hosts to be more promiscuous, the parasites will be more likely to make it to a new host. It is unclear if there are few examples of this type of manipulation because it is understudied or because it is uncommon among directly transmitted parasites.

Now that you know what adaptive manipulation is and isn’t and what types of manipulation exist, you’re ready to go explore the many amazing examples that scientists have discovered or maybe to go discover a new example of your own! If you’re looking for good places to start, I recommend this TED talk by Ed Yong and this cartoon by The Oatmeal.


Moore, J. Parasites and the Behavior of Animals. (Oxford University Press, 2002).

Poulin, R. Chapter 5 – Parasite Manipulation of Host Behavior: An Update and Frequently Asked Questions. in Advances in the Study of Behavior (eds. Brockmann, H. J. et al.) vol. 41 151–186 (Academic Press, 2010).

Parasitic Castrators: Population-level Effects

For background info about parasitic castrators, check out the previous two posts, here and here.

As promised, this week, I’m going to talk about the effect(s) of parasitic castrators on host populations.  I’m specifically going to talk about trematode parasites (Halipegus occidualis) that castrate their snail hosts.  As a motivating statistic, 60% of snails in Charlie’s Pond (the pond studied in Negovetich and Esch 2008) may be castrated by H. occidualis trematodes.  That’s 60% of the snail population that can’t reproduce!!!  It would be hard to imagine castration at that scale NOT having an effect on snail population dynamics.

Whenever math and snails happen in the same paper, I am very happy.  In this paper, Negovetich and Esch (2008) used size-structured population matrix models to look at snail population growth.  They took parameter estimates for the size-specific growth rates, fecundity, and survival rates of snails from a previous field study.  They also took the natural prevalence of trematode infection (60%) and the natural yearly probability that overwintered snails cleared their parasite infections (35%) from the previous field study.

They calculated the natural snail population growth rate (λ) by finding the dominant eigenvalue of their annual matrix, just like you’d do for any matrix model.  Using the natural trematode prevalence and clearing rate, λ was 1.55.  They also calculated a hypothetical snail population growth rate for when there was 0% trematode infection – in that case, λ = 2.41.  So, having the natural levels of trematode infection resulted in a 36% decrease in snail population growth rates!!


I’d vote for Gary.

Negovetich and Esch (2008) also used sensitivity and elasticity analysis to see how changing the various rate parameters – fecundity, survival, growth – would affect the snail population growth rate.  They found that decreasing the size that snails reached reproductive maturity was the most effective way to increase the snail population growth rate.  You might remember from the previous posts that decreasing the size at reproductive maturity is one way to get fecundity compensation.

So, here’s where I quibble with the authors a bit.  They point out that that decreased size at reproductive maturity (=fecundity compensation) has been observed in several systems with snails and castrating trematodes.  Very cool!  But Negovetich and Esch (2008) also imply that these decreases are destined to evolve because they’re the best way to increase snail population growth rates according to this model.  And maybe if snails don’t do something to increase their population growth rate, the large-scale parasitic castration will lead to extinction.  I don’t like this explanation because selection acts at the individual level – the snail population isn’t deciding as a whole to reproduce at smaller sizes.

As we’ve talked about previously, individuals may reproduce earlier when they’re infected either because they know they’re about to be castrated or because the parasite’s manipulation of host resource allocation causes the snails to reach reproductive maturity sooner than the snails normally would.  There’s clearly some fascinating future work to be had on this topic – what’s the mechanism behind earlier snail reproduction in the presence of castrating trematodes?!  I MUST KNOW.

Until then, check out this paper!


Negovetich, N.J., and G.W. Esch. 2008. Quantitative Estimation of the Cost of Parasitic Castration in a Helisoma anceps Population Using a Matrix Population Model.  Journal of Parasitology, 94(5):1022-30.

Parasitic Castrators: A Model

Last week, I introduced you guys to the wonderful world of parasites that castrate their hosts.  This week, I want to talk about the mechanisms behind parasitic castration – both how and why castration happens – and how the wacky responses (e.g., gigantism) of castrated hosts emerge naturally from these mechanisms.  So, enter, from stage left, a beeeeeeeautiful paper by Hall et al. (2007).

Pattern Oriented Modeling:

The Hall et al. (2007) paper is a wonderful example of pattern oriented modeling. Actually, pattern oriented modeling – as championed by Grimm, Railsback, and colleagues – is an approach used to make agent based models more rigorous and more ‘applied.’  For ABMs, using just one predicted pattern isn’t a very useful way to build a mechanistic model, because many different agent behaviors might result in the emergence of the same pattern.  How would you figure out which behaviors were mechanistically appropriate?  If you instead have several expected patterns, it is unlikely that just any agent behaviors will cause all of the patterns to emerge.  So, you can determine which mechanisms are biologically appropriate by determining which agent behaviors lead to your patterns.  Anyways, Hall et al. (2007) used differential equation models, instead of ABMs, but they still used a pattern oriented modeling approach that was very powerful.  Specifically, they knew that any model with parasitic castrators in it should produce the following patterns:

  1. Gigantism.  At any given host age, hosts that are parasitized by castrators should be bigger than unparasitized hosts and hosts parasitized by noncastrating parasites.
  2. Life span.  Hosts that are parasitized by castrators should live longer than hosts parasitized by noncastrating parasites.
  3. Fecundity compensation.  Hosts that are parasitized by castrators should begin reproducing sooner than hosts parasitized by noncastrating parasites.
  4. Reproductive rates.  Hosts that are parasitized by castrators should have lower mean reproductive rates than hosts parasitized by noncastrating parasites because castrated hosts are sterilized, either partially or completely.
  5. Parasite reproduction.  Hosts that are parasitized by castrators should produce more new parasites than hosts parasitized by noncastrating parasites.
  6. [There’s a sixth one in the paper, but I’m going to let you guys go check it out.]

Not only are these patterns predicted from theory, but they had also been demonstrated empirically in a system where bacterial parasites castrate Daphnia hosts.  So, Hall et al.’s (2007) model is based on that Daphnia-bacteria system.

The Kooijman DEB model

Before thinking about parasites, Hall et al. (2007) adapted a dynamic energy budget model to the Daphnia-bacteria system (or a similar system).  In a nutshell, hosts acquire, assimilate, and store energy from their food, and then they can use that energy for maintenance of their current state, growth, development, or reproduction.  Specifically, hosts devote a fixed proportion (k) of their energy reserves to growth and ‘structural maintenance.’  They spend the other fraction of their energy (1-k) on other stuff.  If they aren’t reproductively mature yet, they spend the (1-k) on development, and if they are mature, they spend the (1-k) on reproduction and ‘maturity maintenance.’  The two maintenance things are a bit confusing, I know, but just remember that there are two types of maintenance, and organisms need to spend some energy maintaining their current condition before they can spend any energy on other things.

Host Starvation:

Using this dynamic energy budget model, if the host starts to experience ‘moderate’ starvation conditions, its reserve energy stores will decline. Eventually, the reserves will be low enough that the proportion of energy available for growth and maintenance (k) will only be enough to cover ‘structural maintenance.’  The host doesn’t have any energy left in the ‘k’ part of the budget to devote to growth.  The host still divides up the (1-k) part of the budget to development or reproduction and ‘maturity maintenance,’ depending on whether the host is reproductively mature or not.  So, at this point, the host is still reproducing or developing.

However, if starvation conditions become severe, the host might not have enough energy to devote to reproduction and development.  The host might just spend all of its k and (1-k) energy budgets on maintenance.  And if reserves get so low that the energy stores don’t even cover the maintenance costs, the host dies.

Consumer Parasite:

Now let’s add a parasite to the model!  First, let’s add a noncastrating parasite, which Hall et al. (2007) called a ‘consumer’ parasite.  The parasite takes energy from the reserve energy stores of the host.  In that way, the host and parasite are competing – they’re both using and budgeting that energy pie.

Importantly, consumer parasites can affect how hosts allocate energy, but only indirectly.  That is, if the parasite uses enough of the reserve energy that the host finds itself in moderate starvation conditions, the host stops growing.  It devotes all of the k fraction of the energy to structural maintenance.  And if the parasite uses up enough of the reserve energy that the host finds itself in severe starvation conditions, the host stops developing or reproducing.

Hall et al. (2007) also added one more assumption – if the parasite takes up more than a certain proportion of the host’s total volume (p), the host dies.  Otherwise, very efficient parasites can reach unrealistically huge population sizes.

Ok, sooooo – what happens when we run the consumer model?  After several days of normal growth and reproduction rates, the hosts stop growing because the parasites have reduced the energy stores to moderate starvation conditions.  The hosts keep reproducing for a few more days, but reproductive rates start to decline because the (1-k) piece of pie gets smaller and smaller as the parasite populations grow bigger and bigger.  Finally, the parasites cross that critical volumetric proportion (p) and the hosts die.

Parasitic Castrators:

Now let’s add a parasitic castrator to the model, instead of a consumer parasite.  To do this, Hall et al. (2007) changed just one model assumption.  Instead of leaving k as the constant proportion of energy devoted to growth and reproduction, now k is a dynamic variable.  Specifically, k is a function of the number of parasites per host.  Whereas consumer parasites could only indirectly affect host’s allocation of resources, castrating parasites can directly manipulate host’s allocation of resources by directly affecting k.  (They can also indirectly affect host resource allocation, just like consumer parasites.)  When there are no parasites, the fraction of energy devoted to growth and maintenance (k(0)) is regular old k, from the baseline model.  But as the number of parasites increases, k also increases monotonically – the host spends more and more of the energy budget on growth because castrating parasites directly cause the hosts to budget energy differently.  And because k increases with parasite density, the energy available for development or reproduction (1-k) decline with parasite density.  So, the host is castrated.  [Insert “ooooooohs” and “ahhhhhhhs” of wonder here.]

Ok, sooooo – what happens when we run the castrator model?!  For just a few days, host growth and reproduction rates are normal because parasite populations are still too small to be consuming a lot of host energy.  Then, as parasite density increases, k starts increasing, and host growth rates increase dramatically.  In fact, castrated hosts get much bigger than uninfected hosts and hosts infected by consumer parasites.  After just a few more days, hosts stop reproducing because (1-k) is so low.  Hosts keep growing until just before they die, and host death occurs later than it did when the parasite was a consumer instead of a castrator.

Some Cool Notes:

As we expected, hosts infected by castrating parasites ended up having lower reproductive rates (Prediction 4).  But did they see fecundity compensation (Prediction 3)?  It’s hard to tell from the graphs, but Hall et al. (2007) said that hosts infected by castrating parasites started reproducing sooner than snails infected by consumer parasites and uninfected hosts.  The really cool thing is that the earlier reproduction was completely emergent – it wasn’t “programmed” into the model.  Because castrating parasites increased the growth rates of the hosts, the hosts reached the reproductive maturity threshold size sooner, and thus switched from spending their (1-k) energy budget on development to spending it on reproduction earlier than other hosts did.   Super cool!

Under these conditions, being a parasitic castrator was a better strategy than being a consumer parasite because castrators ended up producing more new parasites than consumers. Consumers rapidly used up all of the available host energy reserves, while castrators kind of “stocked up” energy by getting hosts to grow bigger.  (Bigger hosts can acquire and assimilate more energy.)

Back to Patterns:

Let’s go back and see if all of the predicted patterns emerged from these models.

  1. Gigantism.  CHECK!
  2. Life span.  CHECK!
  3. Fecundity compensation.  CHECK!
  4. Reproductive rates.  CHECK!
  5. Parasite reproduction.  CHECK!
  6. NOT check! In the discussion, Hall et al. (2007) note that their sixth pattern didn’t emerge from the model.  So, there’s some future work to be had there.  It doesn’t mean that this model is bad, but there is some interesting biology and math that needs to be considered further.

So… that was flipping awesome.  In case you’re wondering, the correct thing for you do right now, if you haven’t yet, is to follow this link to the PDF of the paper.  Happy reading!

Fecundity Compensation


Hall, S.R., C. Becker, C.E. Caceres. 2007. Parasitic castration: a perspective from a model of dynamic energy budgets. Integrative and Comparative Biology 47(2): 295-309.

Parasitic Castrators: Introduction

A few weeks ago, I wrote a post that explained the differences between predators, parasites, and parasitoids.  I included a classification scheme that further divides parasites up into smaller groups, including a group called “parasitic castrators.”  For the next few weeks, I’m going to do a miniseries focused on parasitic castrators, because they’re awesome.  Most of today’s post is based on a review by Lafferty and Kuris (2009).

What is parasitic castration?

Rather than reinventing the wheel, I’m going to give you some published definitions for parasitic castration:

Lafferty and Kuris (2009): “an infectious strategy that requires the eventual intensity-independent elimination of host reproduction as the primary means of acquiring energy.”

Baudoin (1975): ‘‘a destruction or alteration of gonad tissue, reproductive behavior, hormonal balance, or other modification that results in reduction in host reproduction above and beyond that which results from nonselective use of host energy reserves by the parasite.”

So, any parasite might reduce host reproduction.  For instance, the host may be temporally too sick to reproduce.  But parasitic castrators specifically target – usually completely eliminate – host reproduction.  Parasitic castrators don’t necessarily immediately sterilize hosts, however.  It might take some time for the infection to mature to the point of complete castration.  And hosts aren’t always permanently castrated – they may regain the ability to reproduce after the infection runs its course, if it ever does.

Stealing Host Energy:   

Castrator energy budgets

I like to think about this in terms of pie, because, well, pie is awesome.  Imagine that hosts have some amount of energy that they have acquired by eating stuff.  Hosts have to divide that energy pie among all of their various requirements – maintenance of their current condition, growth, development, reproduction, etc.  For simplicity, let’s just divide the energy pie into energy spent on reproduction and energy spent on “other stuff.”

When hosts are infected by a non-castrating parasite, that parasite ‘steals’ some of the host’s energy pie.  Non-castrating parasites don’t directly affect how the host divides up the pie; they just affect how much total energy is in the pie.  So, we can imagine the pie being divided up the same way, but shrinking.  (Next week, I’ll talk about a really cool modeling paper that explains how non-castrating parasites can indirectly alter the way the pie is divided up.)  Obviously, when the total pie shrinks, the “reproduction” piece of the pie shrinks, too.  The non-castrating parasite may thus reduce host reproductive output without actually targeting the energy that the host spends on reproduction.

Parasitic castrators directly affect how hosts allocate their resources.  Specifically, they stop hosts from spending any energy on reproduction.  One way to think about this is to imagine that all hosts (uninfected or infected) start with the same energy pie, but then parasitic castrators eat all of the energy that hosts would normally spend on reproduction (bottom left pie).  In that case, infected hosts still end up with a smaller overall pie (like what happened with non-constrating parasites), but this time the whole pie is devoted to “other stuff.”

Mechanistically, I don’t think it makes sense to say that castrating parasites only eat as much energy as the hosts would normally devote to reproduction.  (I can’t tell if Lafferty and Kuris would agree with me or not.)  Instead, we’ll say that castrating parasites change the way that the energy pie is divided up, so that none of the energy is spent on reproduction, and then castrating parasites eat some proportion of that pie.   And that proportion may or may not be equal to the proportion that the host would spend on reproduction, if it had the option.

One more point:  for non-castrating parasites, the host’s energy pie always gets smaller.  But for castrating parasites, the energy pie might actually get bigger!  If hosts are no longer spending energy on reproduction, they’re no longer spending any time with behaviors associated with reproduction – finding mates, for instance.  They might just spend that ‘free time’ resting.  But they might also spend that time foraging, which would lead to increased food intake and thus increased total energy in comparison to the energy pie of an uninfected host.  There aren’t any generalizations about when the energy pie should shrink or grow (that I know of), but it probably depends on the host-parasite system and how long the host has been infected.  That would be a cool modeling paper!

Who are the parasites?  Who are the hosts?

Now that I’ve explained what parasitic castrators are, theoretically, I’ll give you some examples.  Trematodes that castrate their mollusk hosts are probably the most well known example, and they are, of course, my personal favorite.  In this case, mollusks are the first intermediate hosts of the trematode parasites.  There’s a cool paper by Hechinger et al. (2009) that found that 13-39% of the “snail tissue” in first intermediate hosts for trematodes was actually trematode tissue!  That’s a lot of parasite!

You might be wondering if humans have any castrating parasites.  The answer is (please, Evolution, don’t try to prove me wrong): nope.  Lafferty and Kuris (2009) argue that our tiny ovaries and testes are probably just too small to be worth eating.  Good hosts for castrators usually need to have a relatively large amount of reproductive tissue and they need to live long enough for it to be worth the parasites’ trouble to manipulate them.  If the host doesn’t have enough reproductive tissue or doesn’t live long enough, it’s probably more beneficial to just eat the whole host, like a parasitoid does.

Here are just a few more examples:

I have previously blogged about Sacculina barnancles and their crab hosts.

Strepsiptera castrate their insect hosts.

The Extended Host Phenotype:

If you’ve followed any of the work about “zombie” hosts, you might have heard of the extended host phenotype.  Lafferty and Kuris (2009) argue that when a host is infected by parasites that manipulate host behavior, it might still look like a host, but it’s actually just a parasite vehicle.  (Just like any organism may just be a fancy vehicle for DNA?)  In fact, they call parasites that manipulate hosts body snatchers, because that’s what the parasites do – they snatch host bodies.

So, the extended host phenotype is the phenotype of the infected host.  And depending on what parasite the host is infected with, the phenotype may change a lot or a little with infection.  With castrating parasites, one common phenotypic change is host gigantism: hosts get much bigger than they would if they were uninfected.  Note that host gigantism doesn’t happen in all systems with parasitic castrators, but it does happen in some systems.

Another phenotypic change that is sometimes associated with parasitic castrators is earlier reproduction by infected hosts.  This is called fecundity compensation.  The common explanation is that if hosts can tell that they’ve just been infected by a castrating parasite, they should quickly produce as many babies as they can before they become castrated.  (I’ll talk more about fecundity compensation next week.)

Finally, you might be wondering how the parasite manipulates the host.  Parasites can manipulate the allocation of host resources by manipulating host hormones.  And Lafferty and Kuris (2009) point out that parasitic castrators might be forced to specialize on hosts in order to evolve just the right tools for host manipulation.


How do parasitic castrators affect host populations?

So far, we’ve only talked about how parasitic castrators affect individual hosts.  But if individual hosts are being castrated, we might expect host population densities to decline because there are fewer hosts reproducing.  And we also expect that the higher the prevalence of infection (and thus the more castrated hosts), the more the host population density should decline.  This is a research topic ripe for plundering!  Now, I don’t want to put words in your mouth, but you might be thinking, “Wow, this population-level stuff would make a fascinating blog post!”  If you’re thinking that, come back in two weeks!


Hechinger, R., K.D. Lafferty, F.T. Mancini III, R.R. Warner, and A.M. Kuris. 2009. How large is the hand in the puppet? Ecological and evolutionary factors affecting body mass of 15 trematode parasitic castrators in their snail host. Evolutionary Ecology, 23(5): 651-667.

Lafferty, K.D., and A.M. Kuris. 2009. Parasitic castration: the evolution and ecology of body snatchers. Trends in Parasitology 25(12): 564-572.