Wednesday, March 1, 2017

Wolbachia the Holy Grail - Now go away or I will taunt you a second time

*update: Seth Bordenstein let me know that they DID codon optimize their constructs. The text below reflects this change.
**update, part 2: below I note some comments (in blue) as a result of interactions with lead author John Beckmann via the twitter. He was kind enough to participate in a long public and then private exchange - thanks to him.

This morning I was pleased to see that John Beckmann's manuscript describing some enzymatic functional activity of candidate CI genes was just made available on the Nature Microbiology page.  So, today, we will dive again into the world of CI-inducing loci and explore what this manuscript can tell us about the mechanisms behind CI and what these two proteins might do in the insects.

Beckmann et al start out with a different premise from that of Le Page et al. in that John had already identified two candidates, wPa_0282 and wPa_0283, from another study, which focused on proteins found in Wolbachia-modified mosquito sperm. They state that these proteins must have been "secreted into germline cells by Wolbachia." but to be honest, that is not clear from the data. Yes, the proteins may have gotten to the germline but we don't yet know the mechanism.

**In reply, on his blog, the lead author states he "never explicitly say[sic] it is secreted".  This text above was taken directly from the Beckmann et al paper, in the abstract, so unless your interpretation of "secreted by Wolbachia" is very different from mine, this means something specific. After our back and forth, John says that although he is ultimately interested in how this happens, he is not focusing on it for now. 

Just like Le Page et al., Beckmann et al highlight the fact that these two loci occur in syntenic orientation with respect to one another. wPa_0282 (which they name cidA) is directly upstream of wPa_0283 (named cidB) and this is true across homologs, as published by Le Page et al.  Beckmann et al take the stance that this orientation, and conservation of orientation, must mean these genes are co-transcribed, in an operon. I am not so sure that there is very strong evidence of these being in an operon -- that said, I am not sure that matters much here, except for pushing the idea that this is a toxin-antitoxin system. But in that case, one should show that the antitoxin is less stable (we'll get to that later).

In Figure 1 of the paper, below, they highlight three orthologs, two in wPip and one in wMel, which they work on throughout the rest of the manuscript.  Interestingly, the cidB gene harbors a Ulp1 protease domain which is missing from its homolog within wPip, cinB.  




Based on the analysis of these three homologs the statement is made that: “As Wolbachia strains evolve within different host species, they accumulate mutations in their corresponding CI systems and become bidirectionally incompatible.”  and also “The evidence to date is most consistent with a duplication and divergence from a common CinB-like ancestral operon, although the converse cannot be ruled out, namely, that an ancestral CidB-like operon picked up a nuclease domain and later lost the DUB domain." These are testable hypotheses, sure, but certainly not shown by the reference cited nor really by the data provided in the manuscript, although in the Supplementary Discussion, they do describe some of the variants and homologous loci in other Wolbachia.  In this paper, they show that cognate cidA-cidB proteins interact in vitro and that cinB will not interact with CidA variants (wMel or wPip), but this is as far as the pattern goes with regards to this dataset.  Again, it will be cool to explore this further with the different variants out there.

**One of my pet peeves is over-interpretation of data or broad sweeping claims without substantiated support and I tend to come down hard on folks for it, in review and in post pub review.  The authors may be working on these as we speak but the data were not in this manuscript.

So, next comes the real meat of the paper, identifying DUB activity for the CidB protein.  One thing that is awesome is that they have made catalytically inactive mutants by altering the residues known to be important for the DUBs and the nuclease. The beginning of this work is done in yeast, expressing the proteins and looking for growth defects (as we've done with Wolbachia effector WalE1). One thing I would've liked to have seen in their yeast work is a graphic of constructs.  Are they really expressing these as an operon in yeast as well? That wasn't super clear to me from the manuscript, but maybe I missed it.

**I was confused as to which constructs were used when because of the following statement in the methods section of Beckmann et al : "The 2-micron plasmids pYES2 (URA3) and p425GAL (LEU2) both had the GAL1 promoter and CYC1 terminator and were used for galactose-induced expression of Wolbachia genes in yeast37. Expression from the low-copy CEN vector pRS416GAL1 was also used."  It's important to know if things were expressed from low or high copy vectors, because that will alter the amount of protein produced and could explain why they still see a growth defect when CidA+CidB are expressed.  If we know something about how these proteins interact with each other, and at what ratio, it could let us know something about function. In a private exchange, John clarified that protein expression was controlled in their study and that due to space constraints in Nature, they could not put much in the figure legends (limit to 100 words).

 CidB in yeast induces a conditional growth defect - only at 37C do they see growth inhibition (Figure below).   Check out the Galactose + 37C for CidB (fourth lane from top in panel a).



What's cool here is that they can take advantage of their catalytically inactive mutant (C1025A) and show that it does not induce the growth defect so definitely the Ulp1-protease activity is behind the yeast phenotype.  Also, co-expression of CidA, the upstream gene, rescued the phenotype to some extent (the yeast cells still look plenty unhappy and probably if you compared the CidB+CidA at 37C yeast to the Vector 37C yeast you'd see a significant growth defect).  They had a harder time identifying the exact substrate in yeast, as the profile for Ub-conjugated proteins upon CidB expression did not dramatically change.

**The lead author has a lengthy explanation of this in his blog, which you should read if you are interested to know the details.


That said, I am sure they are taking advantage of the yeast deletion library to identify candidates for both synthetic rescue and lethality.  Regardless, they do identify one substrate for CidB (UbVME), which they use in their in vitro assays (see below).

**The lead author has a problem me calling UbVME a substrate, because it is not a real substrate for DUBs that would ever be found in the cell. Read more on his blog for that. That said, what they are effectively doing below is an in vitro assay testing for DUB activity, using a byproduct they can measure, and a known substrate/target (UbVME).



Riffing on the idea that this is a toxin-antitoxin system, they go on to try and explore if CidA inhibits CidB activity in vitro (see panel (a) above).  It does not.  So, how do we explain that CidA expression seems to rescue CidB expression in yeast? Well, I'd love to know what these constructs look like again - is expression is the same for CidB when you've got the CidA construct around?  Are these plasmids kept at the same copy number?  They suspect that perhaps CidA rescues CidB toxicity by changing localization of CidB. Why not look at localization in yeast? See if that changes with CidA expression? Lots of questions about this one.

**John has data on this! cool. More to come from this group will be exciting to see.

Moving on to the final part of the study - showing that expression causes CI in flies.  They go about making their constructs in a different way from Le Page et al. First, they codon optimize their constructs (this is not mentioned in Le Page et al so I am guessing they did not this was also done by Le Page et al).  Second, like the black night that never dies ("It's just a flesh wound!") the hypothesis that these loci are in an operon persists, so they do this interesting trick with T2A to make sure that these syntenic genes would be expressed and translated as two different proteins.  Why the obsession with it being an operon. Honestly, I don’t really get it.  Many genes encoding proteins that function together in Wolbachia are NOT found in operons (T4SS exists in multiple operons, for example), and it does not mean they are not important.  

**The fact that effort was put into making these constructs in the fly as a pseudo-operon is interesting to me. Does this change how they function? Why would it? This is a completely artificial system (transgenic fly expressing wPip genes) - and this the lead author agrees with.  Whether they are in an operon or not, these are interesting loci.  I am not sure if Beckman et al is working on this or not, but as both of these transgenic flies are now published, someone should (not me!)

Although flies expressing these genes are fine, they get severe lethality here when they cross males with uninfected females.  You might notice if you compare Figure 3 of the Le Page manuscript, that expression in the system used by Le Page is no where near as severe.  It's very cool that they could again eliminate the sterility by changing the DUB catalytic residue (C1025A).  It's also nice that, like Le Page et al., they can show the embryonic developmental abnormalities when you express cidA-cidB in flies (figure below).

With regards to the severity of the lethality observed, this could be either 1) expression levels depending on the insertion point or 2) using codon optimized constructs while Le Page et al did not,  3) the mosquito homologs are more lethal or 4) 3) something interesting here about the "operon" construct.  As they did not show strength of expression in the fly it is hard to say at this point - would be interesting to compare the transgenic strains from LePage to these.

**No real reply from John about this - maybe he has something else in the works.

Finally, unlike Le Page, they could not isolate a strain singly transgenic for cidB, suggesting it is toxic in the fly.  Also, they do not show rescue, either with a Wolbachia infection (because they don't have wPip in Drosophila melanogaster) or with cidA or cidA-cidB expression (which suggests that cidA is not really a rescue factor).

**This is important.  Co-expression in the yeast system DID alleviate lethality - and this result cannot be ignored.  However, are we observing artifacts of the expression system (either yeast or fly?) Does this result depend on protein expression levels (I imagine it would).  Why is cidA required for CI in flies if it is the "rescue"? The lead author has some interesting hypotheses here - perhaps cidA is cidB's chaperone.  Perhaps it regulates something about secretion in the natural system.  Perhaps one or both of these proteins are modified in a way that is not possible in the transgenic system.  In the naturally infected ovarian proteome, the cidA protein is extremely abundant, so maybe expression levels matter.  Looking forward to more from both labs on this.

There's a long supplementary discussion, which I won't go into, in which they try to argue that this result does not kill off the "antidote-toxin" hypothesis ("what are you going to do, bleed on me?" "I'm invincible!").  I agree with Beckmann et al that the fact that these two loci are syntenic, and potentially in an operon, is intriguing.  Is there any evidence that CidA (the hypothesized antidote) is less stable than CidB?  They show in supplementary figure 2 that E. coli Lon protease can cleave CidA, but Wolbachia Lon does not seem to.  No matter, if we believe these are acting in the eukaryotic cell, then wouldn't a eukaryotic protease be involved? Is there any evidence from their yeast system, for example, that CidA is less stable? These are interesting questions to pursue in future, as is whether the type IV secretion system actually secretes these or if transfer is facilitated by another mechanism (perhaps WO).


**To pull a quote from John Beckmann's blog here "Specifically testing secretion via type IV system is a waste of my time". Well, being that I am deeply interested in Wolbachia type IV secretion, mechanisms of host interaction, and host/wolb evolution, I seriously disagree. It may be a waste of John's time, but it is certainly important for everyone in the field to establish what is secreted by Wolbachia when and how these effectors modify host processes. Shouldn't it matter if instead, as suggested by Le Page et al., these are part of the phage? It certainly means something very different for Wolbachia.










Tuesday, February 28, 2017

The Wolbachia "holy grail"

The once-obscure alpha-proteobacterium Wolbachia pipientis, was catapulted into medical relevance by the discovery that it inhibits the replication of RNA viruses.  However, it was first discovered for its odd quirk of manipulating insect reproduction. That's right, Wolbachia were called "reproductive parasites" and for a long time, no fitness benefit was attributed to them.  These intracellular masters do things like kill male offspring, feminize male offspring, induce parthenogenesis, and the most common phenotype, sperm-egg incompatibility (also known by its more complicated name: cytoplasmic incompatibility or CI) (Figure 1).  



Figure 1. Modified from Werren et al., 2008.  Wolbachia cause four distinct reproductive phenotypes in a range of arthropod orders (top). 

The mechanisms behind CI - meaning how does Wolbachia induce it - have been a mystery for some time.  When you make a cross between infected males and uninfected Nasonia females, unviable embryos result from a mistiming of nuclear envelope breakdown for the male and female pronuclei, suggesting some cell biology behind the failed crosses. With regards to the Wolbachia genes facilitating it, however, there are so many papers on potential models that explain the data - from lock/key to mod/rescue - in fact, when you search google scholar for these terms, you get 92 and 281 papers, respectively.  Folks have mined the existing Wolbachia genomes for potential candidate loci in many different analyses, without any substantial evidence for loci underpinning the interaction.  I, for one, had always been skeptical of these types of approaches because - is Wolbachia evolution so static as to only use one homolog or one mechanism for such a widespread phenomenon. 

So, back in June of 2016, when the Wolbachia meetings took place in Australia, two research groups presented data on two loci - WD0631 and WD0632.  It is with great anticipation that the Wolbachia community awaited the recent manuscript from both John Beckmann and the Bordenstein lab, which I review openly here.

The paper starts out with the premise of a bioinformatics search to identify CI candidate loci. They use the following criteria: present in all CI-inducing strains, and absent or diverged in non-CI strains, expressed in the gonads of infected insects (especially in the C. pipiens ovaries).  Amazingly, this list led them to only two loci - WD0631 and WD0632 (the wMel homologs).  These two loci are associated with Phage WO, the lambda of the Wolbachia world, and are actually be part of the "Eukaryotic Association Module" which is packaged into phage particles. This means that these loci are actually phage loci, not strictly Wolbachia loci.

Now, what other characteristics would we expect from CI-inducing proteins? Well, they would be highly expressed in young male flies (as these are the males for which the CI phenotype is most penetrant).  Are these two loci differentially expressed in young males? Yes, see Figure 2 below:


Figure 2. Expression (qPCR, normalized to Wolbachia groEL) for the two CI loci (WD0631 and WD0632) is upregulated in young male flies, as is the expression of other phage associated genes (WD0508, WD0640 and WD0625). A non-phage associated gene (WD0034) is not upregulated in the same way.

Yes, WD0631 and WD0632 seem to be more expressed in young males but interestingly, so are other loci here.  The uniting factor? These are phage associated...curiouser and curiouser.

What else would we expect of these loci if their protein production was actually behind CI? Well, we would expect to find that you could recapitulate CI by expressing the proteins in transgenic flies.   LePage et al do this (Figure 3b below)! Check out the cross between uninfected (unfilled) males expressing WD0631 and WD0632 and infected females AND the rescue between infected females and those same males.



Figure 3. Expression of CI genes in the reproductive tract of male flies recapitulates CI and is rescued by Wolbachia infected females - CI measured by counting hatched embryos.

And in the most convincing figure of this set (in my opinion) they show a dosage dependence of sorts by exacerbating CI in Wolbachia infected males by expressing the transgenes either alone or together (Figure 3c above).  What's interesting is that with one trangene alone, they could not get a phenotype (Figure 3A), suggesting that these two proteins work synergistically in some fashion.

To add the cherry on top, LePage et al look at embryonic defects in these transgenic crosses and find that, just as in "real" CI, when you express WD0631 and WD0632, you see embryonic defects such as chromatin bridging and regional mitotic failure. Importantly, these defects go away when you mate transgenic males expressing the CI loci to infected females! Bingo! They have satisfied all of the requirements necessary to show that these proteins are indeed involved in CI. 

But how do these proteins induce CI? How are they transferred to the host (by phage???!!!) and what do they do in the host cytoplasm? Do they function together? Are there other loci that underpin CI or other reproductive manipulations? Some of these points may be, in part, addressed by John Beckmann's mechanistic work.  I am sure there will be more interesting work to come from this group!