Sunday, 22 April 2012

What is this 'effector' thing, anyway?

In which I opine about the definition of (plant) pathogen effectors.

I took part in a Twitter discussion between plant pathologists recently, attempting to reach some kind of agreement on what an 'effector' is.  It's hard to make some points coherently in a 140-word limit, but in the end it seemed that everyone went away thinking that an understanding had been reached.  There were some things in there that I wasn't entirely happy with - I found it very difficult to be precise - so I thought I might expand a bit, here.

A quick intro for non-pathologists

Plants have an immune system that protects them (much of the time) from the onset or spread of disease, but it doesn't work quite like ours.  There's no real reason why it should - we diverged from plants a long time ago, and plants are essentially immobile: they don't have the opportunity to move away from danger.  Like the human immune system though, plants raise chemical defences against microbial invaders, and they need to assess the level of threat, convert it to a signal, and raise an appropriate level of response.  Too strong a response could be wasteful of resources and limiting of growth; too weak a response could leave the plant vulnerable.

MAMPs For The Memories

Picture (if you can) a microbial pathogen approaching a plant cell: a fairly flexible bag of living chemistry possibly thrusting itself through a fluid, or along a leaf (or root, or stem...) surface.  As it goes, it is likely to be producing and secreting a number of molecules that aren't produced by plants.  It is probably also shedding parts of itself, in the form of larger molecules, and fragments of larger molecules.  There is, then, a fairly sparse molecular cloud of non-plant chemistry accompanying the pathogen as it approaches the plant cell, which looms like a large shed (plant cells tend to be a bit more solidly-structured than pathogen cells).

The plant cell is not sitting there, entirely inert.  The outer surfaces of its cell walls are 'hairy' with detectors.  These are triggered by chemicals in the surrounding environment and are the eyes and ears of the plant cell.  The plant is also producing enzymes and pumping them into the surrounding locality: these can break down larger molecules in the vicinity into (more) smaller chemicals, which can then be detected more easily.  The overall effect is a bit like a chemical soup.

When the plant cell's surface detectors come across something unusually 'non-self' floating in the immediate chemical soup, this can be taken as a sign of imminent invasion or infection, and signals can propagate through the plant cell to raise up the appropriate defence.  The elements that come from an invading microbe that float through the soup are known as Microbe- (or Pathogen-) Associated Molecular Patterns: MAMPs or PAMPs.  The terms really only differ depending on whether we want to highlight the microbial or pathogenic nature of the organism being detected.

The Effects of Effectors

The invading microbe isn't usually passive, either, especially during interaction with the plant.  Microbes that interact with plants are variously able to interfere with the plant's normal operation to reduce the effectiveness of host defences, and to turn the plant's own chemistry to the microbe's ends.  They can do this by introducing into plant cells their own chemical species that either interfere with, replace, or otherwise subvert the biochemical systems of the host.  A number of strategies for introducing these species have evolved, including direct injection through a syringe-like structure (e.g. bacterial type III secretion).  These chemical species are typically what are referred to as 'effectors'.

So what's the issue?

Words are not equivalent to variables in computing languages, in the sense that they are not simple containers that just hold concepts.  We internalise the meanings associated with words and symbols in a way we simply don't do when we assign a value of 5 to the variable a, for example.  We can assign a value 3.14 to a variable x, but when we think of 'pi', we naturally associate many other concepts (transcendental numbers, randomness, circles, trigonometry and so on) and become psychologically 'primed' for those concepts.  The same is true when we talk about any other concept, including that of an 'effector' and, if we do not all share similar associations it is easy to talk at cross-purposes.  We need a common understanding of what we talk about when we talk about effectors.

I believe that, happily, the concept of an effector is pretty straightforward and easy to illustrate with some examples from the 'real world'.

Effectors modify plant behaviour.

This seems a natural place to start.  By behaviour I mean any manifestation of 'behaviour' from changes to chemical responses upwards.  We wouldn't be thinking about the subject at all if there hadn't been some noticeable effect on a plant, somewhere.  That's not to say that an effector always modifies plant behaviour.  Sometimes an effector might need the plant to be in a particular state to be able to act.  Sometimes the effector might need to act together with some other molecule (perhaps another effector) to have an effect.  Sometimes the plant may be able to block the effector's action.  But, essentially, the clue is in the name: effector.

Effectors are molecules produced by microbes.

Not everything that has an effect on plant behaviour is a molecule: plants also respond to changes in temperature, the location of a light source, local pH or other ionic concentrations, and mechanical stress or damage, for example.  We can state a definition that effectors should be identifiable molecules, whether small molecules, proteins, or something else.  This does not preclude the action of effectors requiring dimerisation (or multimerisation), or simple association with other molecules.

However, not everything that a microbe produces is a molecule, or simple association of molecules.  For example, bacteria may produce biofilms, which are complex aggregates of the bacteria embedded within a matrix of slimy extracellular polymers.  These biofilms can result in changes to the local environment, to which a plant may in turn respond.  As a complex structure, biofilms do not count as simple associations of readily definable molecules and so do not - by this definition - count as effectors.

Not every molecule derived from a microbe that modifies plant behaviour is an effector.

MAMPs/PAMPs modify plant responses, but they are not effectors.  They are, rather, tell-tale indicators of a microbe's presence.  The difference between a MAMP and an effector is one of function or (if you find this approach easier to follow) agency.  

Often, MAMPs are fragments of larger molecules with no direct functional activity of their own, such as flg22 - a fragment of the flagellin protein - or fragments of chitin (the polymer that constitutes much of fungal cell walls).  Plant cellular behaviour is modified on MAMP detection, but that is not the function of the MAMP.  This is especially so if the MAMP results from the action of the plant on the microbe: there is no aspect of agency on behalf of the pathogen in the production of the MAMP.  This is what I'm aiming for in the distinction between 'produced by' and 'derived from'.

An analogy might be the detection of an aeroplane by a defence system's radar, and the launch of a missile in response.  Here, the designed purpose of the aircraft is neither to reflect radar waves, nor to trigger the missile launch.  Likewise, the evolved 'purpose' of the MAMP is not to trigger a plant defence response.  Hence MAMPs are not effectors.

Modification of plant behaviour may look like normal plant behaviour.  

If the action of an effector is to disrupt normal plant behaviour, that also includes things like disruption of MAMP detection: essentially, restoring 'normal' plant operation when the 'natural' behaviour could be to raise some kind of defence.

Effectors may act within, or outwith the plant cell.

The type III secretion system mentioned above involves the injection of microbial proteins into the host plant cell, and this is how we (or, at least, I) most commonly think of effectors - as species that act within the host's cell.  This doesn't need always to be so.

To extend the aircraft analogy from above, we can imagine a stealth paint coating that absorbs the radar signal, thus preventing detection.  The stealth paint coating would be considered to have the purpose of disrupting the defence the usual chain of signals that would otherwise result in the launch of missile defence.  By virtue of its action on the 'normal' operation of the defence system, it could be considered to act like an effector.  Notably, this action takes place 'in the open', away from the detection system itself.

There is a protein produced by the fungal plant pathogen Cladosporium fulvum that acts like this stealth paint.  This protein, called Avr4, protects the pathogen's cell wall (specifically, the chitin component) against being broken down by a protein produced by the plant (a chitinase).  Here, the plant's chitinase is like an outgoing radar signal; normally it would elicit a chitin fragment from the fungus (the return signal), which would be picked up by a plant surface protein (the radar dish).  By preventing the generation of a chitin fragment, the Avr4 protein essentially absorbs the outgoing signal, disrupting the 'normal' plant detection system.  This happens outside the plant.  By our definition so far, Avr4 is an effector.

Being required for observable modification of plant behaviour is not equivalent to being an effector.

There are two issues here.  Firstly, many molecules and systems that play no direct part in modification of plant behaviour may be essential for microbial well-being - but microbial well-being may be essential for modification of plant behaviour.  Removing all the enzymes of glycolysis from a microbe may prevent microbe-plant interaction, but they are not effectors because of that.  Secondly, some molecules that affect plant behaviour may be redundant, in that they may be deleted or removed without affecting the change to plant behaviour (e.g. Redundant Effector Groups).  Either case is sufficient to show that being an effector and being required for modification of plant behaviour are not equivalent.

Also, effectors may have a host-specific action: they may disrupt behaviour in some plants, but not others.

Avirulence proteins are not equivalent to effectors.

Avirulence proteins are proteins detected by the plant, and whose detection results in plant resistance (or, at least, a defined resistance response) to the microbial invasion.  Many of the best-understood avirulence proteins are effectors but, since there is no obvious requirement that the detected protein be an effector, or that effectors be detected and result in resistance reponses, the two definitions are clearly not equivalent.

Effectors are not restricted to pathogens.

I have tried not to give definitions that lean too heavily on the biology of pathogens, as disease is a concept with a distinctly human bias.  The effect of a disease is - quite obviously - negative.  But there are many associations between plants and microbes that do not result in what we would, as humans, call disease.  An obvious example is the symbiosis of nitrogen-fixing bacteria with plant hosts.  But even these beneficial (or any number of benign) interactions can involve complex molecular interactions between the microbe and host plant, in which the microbe seeks to modify host behaviour using effectors.  

I consider pathogens to lie at one extreme of a continuum of plant-microbe interaction in which the pathogen gains all the benefit of the interaction and the plant none.  At the other end, the plant gains all the benefit, and the pathogen none.  Through the rest of this continuum, which includes many shades of symbiotic interaction, the picture is not so clear, and an interaction that is apparently mutually beneficial or benign may suddenly turn pathogenic if the environment changes (such as the emergence of pathogenesis under stress).

I think this definition of 'effector' applies quite readily beyond plant-microbe systems, too... 

Effectors act directly.

This is the part of my definition that gives me most trouble.  Where we know how effectors work, they tend to interact directly with a host plant molecule.  Where they do not (such as Avr4), their role in disrupting a chain of communication or signalling is typically quite clear, and we can consider such direct disruption to normal information transfer to be direct action.

The picture becomes less clear when proteins (such as Ace1 from Magnaporthe oryzae, or AvrD from Pseudomonas syringae) act to synthesise chemical species that disrupt the host.  By this definition, the synthesised molecule is the effector, and the synthesising proteins are not effectors, because they 'act at a distance', the same way a ribosomal protein is not an effector simply because it plays a role in synthesising effectors.  However, AvrD is (judging from the name), an avirulence protein.  

The picture is also less clear when we think of proteins that might modify the environment of a host cell to disrupt behaviour.  With Avr4 we can make a reasonable claim that it is disrupting the normal operation of a signalling pathway; it is acting precisely.  Modifications to a bulk property of the environment (e.g. ionic concentrations, temperature, viscosity) do not count by this definition, as they are not direct action, but action mediated by, well... the medium.

Effectors need not benefit the microbe.  At least, not right now.

Although it is difficult to think a reason why an effector would persist indefinitely if it never provided any benefit to the invading microbe, since we are always observing a snapshot of evolution, there is no reason to impose an absolute cost or benefit requirement as part of the definition of an effector.

In Conclusion

I hope I have been able to make a case for the following one sentence definition of an effector (with the caveats above):

"An effector is a molecule produced by a microbe that acts directly to disrupt the normal operation of host biochemistry."

Or, more generally:

"An effector is a simple entity produced by system A that acts directly to disrupt the normal operation of system B."

(where 'simple' is defined appropriately for the two systems).

As ever, brickbats, bouquets, and better definitions are welcomed in the comments.


  1. From Kamoun Lab @ TSL at

    This is definitely in line with my favorite definition of effectors as secreted microbial molecules that "alter host-cell and structure". The key aspect in my view is not to get stuck with whether the effect is positive or negative. That was the problem with the somewhat outdated concepts of elicitor, avirulence, toxin etc. They assume a certain effect on the pathogen (good or bad) while the reality is often more complex and often host-genotype dependent.

    The one difficulty I have with what you wrote is the part about complex structures. Most likely all proteins occur in complex with many other molecules. I don't really buy the dichotomy and I don't think there is anything as simple association with other molecules. That's just a convenient way for us to think of our data.

  2. I wouldn't want to staple my colours to a statement like "all proteins occur in complex with many other molecules", but the interior of cells is certainly very crowded, and there is very little chance that an effector, once within the cell, would have an opportunity to 'diffuse freely through a solvent', or engage in any such idealised behaviour. More likely there are a series of not-very-specific local interactions that are easily 'broken', until a more significant interaction (e.g. a 'binding partner' or 'target') is obtained.

    What I'm thinking of is the distinction between a complex involving an effector that can be described - even in in an idealised way - stoichiometrically (i.e. that there are a set prescribed number of elements of species A, B, C... etc. participating in the complex), and one which cannot (e.g. a membrane, biofilm, or flocculation of proteins). The first of these classes is 'simple', and the second 'complex', in my shorthand.

    I haven't got any firm idea of a hard and fast dividing line between the two - at least, not one I can define better than hand-wavy 'crystal' vs 'glass' - but I think it makes sense to keep a definition of effectors to those species that lie in the 'simple' category, when they act as effectors. The property of 'being an effector' is then constrained to something that is a property of the effector - or a well-defined assembly involving the effector - in the act of being an effector, rather than a bulk property of an ill-defined assembly in which the effector participates.

    Is that clearer? I think it brings our positions closer together, even if it doesn't eliminate the distance.

    As a (once, at least) chemist, I do think that we can distinguish qualitatively between levels of complexity in molecular assemblies. We do (or did...) it all the time with polymers and complex mixtures. There are quantitative measures of complexity that can be applied, but I don't know if they would be of any more than marginal practical use in a definition of effectors.

  3. Fist, on the $variable analogy. I fully agree that the concept that we put behind the word "words" is not equivalent to the one behind "variables". My point was that they are still closer than you think. The meaning of words change over time, and that meaning is set by humans. It is true we cannot completely erase one meaning and put in another. It is true that there are associative links. For me, pi also clicks with a certain class of "dessert a la mode". Nevertheless, as humans and scientists, we need to fully understand that we do have certain control over what we call "effectors", and moreover, that we should update its definition (or come up with new terminology) as we get more knowledge about our system.

    To give an example: many oomycete effectors have been observed to have a certain four letter protein motif. Soon, having such motif became one of the key features of oomycete effectors. So, should we a priori exclude a protein that does not have such motif from qualifying as an effector? Or should we update the definition once we find a few? May be people will disagree, but for me the answer is clear - update. I think that the worst we can do is be dogmatic. As a responsible society of scientists, we must discuss and define our agreements and disagreements. And that's what we are doing, yay!

    I think what Sophien is doing is the following: he is proposing a broader definition of effector (which I do not agree with), in part, to free our minds a little bit. So that we can be less dogmatic about what an effector is and what it is not, and in turn, find effectors where we would not suspect them to be. And that is the thought that I fully agree with. If I am misinterpreting his opinion, he can correct me here.


  4. I have no issue with the fact that the meaning of a word may vary with usage over time (one of my favourites is 'nice':, nor with the fact that words are proxies for concepts. We certainly agree, there. I meant rather to stress the point that we may control what we intend to be the meaning of a word, but not what it represents to the reader, or how it is processed (Interestingly, the reader often exercises no conscious control over this, and in fact reading carefully is cognitively expensive: see for example the Stroop and IAT tests We have a very complex relationship with language, and words direct thought as well as represent it.

    That is different from the role of a variable in computer code, which is, ultimately, the content of an address (set of addresses) in memory. The identifier (variable name) we, as humans, read in high-level code is merely a convenience: the choice of this identifier is irrelevant to the running of the code; the same is not true for human language. As we all know, the need for precision in communication is one of the reasons why technical language tends to be full of jargon and defined terms. Like it or not, 'effector' in the plant pathology sense is a jargon word, and we might be better off defining it precisely.

    The question concerning a certain 'four letter motif' (let's not be shy, it's RxLR, isn't it? ;)) is an interesting one. What it boils down to is this: at least one sequence that was known or suspected to be an effector was observed to contain this motif. The motif was demonstrated to be required for translocation of at least one effector sequence into the host. That provided a hypothesis for the role of the sequence motif. That the motif (and its surrounding region) can be transferred to other sequences, enabling their translocation, confirmed this role. These results say nothing about the unconfirmed biological functions of the other sequences in which that motif is (or is not) found. Nor do they say anything about whether that motif is present in all effectors (it isn't). The RxLR motif isn't so much one of the key features of oomycete effectors, as much as one of the few handles we have on identifying candidate effectors from sequence information alone.

    The definition of 'effector' I outline above has no association with the presence or absence of any particular sequence motif, but depends only on its role in the interaction with the host - the behaviour of the molecule. If we define effectors by their behaviour (as we would define a 'transporter' or an 'enzyme'), then anything that displays that behaviour is an effector. I don't think that this is unreasonable. I certainly think it's safer than defining them in terms of unproven assumptions about sequence motifs ;)

  5. @fmartin54 points out on Twitter: "Agreed. Note: Microbes are not the onlyorganisms releasing effectors. Insects do." (!/fmartin1954/status/194814889037004800)

    And he's absolutely right. I focus on microbes because it's what I'm familiar with, but I think the functional definition can potentially extend beyond microbes, and beyond biology to describe a class of interaction between complex systems.