Tuesday, October 29, 2013

In Defense of the Hologenome

In a perspective article published in Zoology, we recently wrote:
"The hologenome concept of evolution specifies that the animal's genome, mitochondria and microbiome are an aggregate of genes that together form a unit of natural selection (Zilber-Rosenberg and Rosenberg, 2008). The evidence motivating this concept spans the essential roles of the microbiome in eukaryotic fitness (McFall-Ngai et al., 2013), including digestion, immunity, olfaction, organ and neuronal development, etc. However, the hologenome concept is controversial because many biologists view the microbiome as extrinsic to the host animal and therefore unable to co-evolve sensu stricto with the host genome."
As an illustration of the controversy, I am posting an email dialogue that we recently had with a prominent expert in evolutionary biology and a postdoc on our paper on the hologenomic basis of speciation. The discourse below is meant to highlight the points and counter-points of the dialogue and perhaps convert a few skeptics.  I am blogging it because I believe it will be a far more effective way of advancing this discourse into the community, rather than just keeping it between a few scientists with limited benefit. 

And for more information, I posted the video from a google+ hangout on the hologenome below that was comprised of several evolutionary biologists and microbial ecologists (link to blog post). After we got the kinks worked out for the video chat, we had a productive one hour chat. 


Dear XXX and YYY,

Thanks again for getting in touch with your commentary.  We  appreciate the opportunity to address your concerns, and we pledge to make this process productive so that we can all learn something in the discourse. You’ve listed a number of issues in this short commentary, but we think that we can demarcate them into four discrete areas to discuss. Do correct us if there’s anything critical that you feel that we’ve missed. 

Rob and I can start by saying that a short Science paper doesn’t do much justice for convincing the community in one shot. Thus, we expect  and welcome this discussion;  we have encouraged such in our preceding blog posts, google hangout discussions with other scientists (skeptics and supporters alike), and our Zoology article.  And there are more experiments to be published, of course. We also think that posting this discourse on our blogs (either anonymously or not) would be helpful so that the broader community can see the pros and cons of the arguments.

1. " The claim that coadapted gut bacteria cause hybrid lethality and speciation in Nasonia requires that bacterial gut assemblages of Nasonia recapitulate their hosts’ phylogeny, are species-specific, and are coadapted to their hosts in the sense that “foreign” microbiota lead to lower fitness."

There are three issues that you question in the above statements. First, the Nasonia gut microbiome recapitulates the hosts’ phylogeny. We demonstrated this pattern in our Evolution paper, this article, and a third set of experiments that we are currently working on.  So we think your commentary here stems from a misinterpretation of what phylosymbiosis is, rather than its lack of evidence. Here’s how we view it operationally. Phylosymbiosis is simply the reconstruction of the host’s phylogeny with a UniFrac inference method used to compare microbial community relationships. Like phylogenies, UniFrac trees are statistical inferences based on the weighted (Fig 2B) and unweighted abundances (Fig 2C) of OTUs. Remarkably, both the weighted and unweighted Unifrac trees are in complete agreement with a phylosymbiotic microbiome.  We suspect you are laser focused on the pie charts, which create the illusion that G and V have very similar communities based on the single, dominant Providencia genus. The individual OTUs are collapsed and simplified in the pie charts into a broader genus category. If we were to reconstruct the tree based on one OTU, like Providencia, you might be correct; however, these widely-accepted UniFrac trees developed by the microbiome field are reconstructed based on all OTUs in the microbial communities. Take for example the simple comparison that the younger species, G and L, share 41% of the OTUs in their communities while the more divergent species, G and V, share  24% of the OTUs in their communities. 

Your second claim is that the Nasonia gut microbiome is not species-specific.  In the same light as above, any one OTU can be present in all of the species but the metric that we use here is the total microbial community. It is species-specific because (i) there are a number of OTUs (besides Providencia) that are specific to each Nasonia species and (ii) the divergent Nasonia genomes select upon the microbiome diversity that can inhabit them. Because we have placed all of the Nasonia in our studies under identical environmental conditions, the variation we see in species’ microbiomes is explainable by gene-microbe interactions. 

Finally, the third claim that foreign bacteria should cause hybrid lethality is not a requirement at all. In fact “foreign” is an unusual term to describe this process. What matters is the ability of the host genome to recognize and thus epistatically interact with the microbiome. So, an OTU could be foreign in one generation, but still be part of the recognizable microbiota in subsequent generations. Perhaps a simple way of looking at this is from a genetic perspective. Like a beneficial gene within species that breaks down in hybrids, we expect that beneficial bacteria or those closely related to the native microbiome in Nasonia should cause lethality. In contrast, an unrecognizable “foreign” bacteria should not cause lethality because it is not part of the of  recognizable microbial community and thus epistasis within species that breaks down in hybrids. Ongoing experiments in our lab currently support these predictions. 

In conclusion, we urge caution because the claim that "none of these are established" or later that "these data...do not meet these criteria" (and throughout the top paragraph of page 3) dismisses widely accepted means of analysis and interpretation for microbial communities. 

2. "“Phylosymbiosis” connotes relative stability of the gut bacterial communities within species."....Brucker and Bordenstein (1) provide no evidence that the gut bacteria of Nasonia are sufficiently constant and species-specific to contribute to speciation."

Some clarification of the phylosymbiosis model is required in our response here because it seems to us that you have extended the term to something that we do not adhere to. As defined in the Science paper, it is “a term...to denote the microbial community relationships that recapitulate the phylogeny of their host”. Phylosymbiosis is a pattern that is not necessarily derived from a continuous process of stable OTUs each generation. Instead, phylosymbiosis derives from the host genome interacting with the environmental microbiome to collect an assemblage of (similar) microbes whose community relationships parallel the evolution of the host genome each time they are measured. We come to this conclusion because the null hypothesis for a totally random or diet-centric collection of microbes in a host is that phylosymbiosis would not be evident in the microbiome data.

Like phylogenomics, phylosymbiosis is the total microbiome metric of the host’s evolutionary        relationships. As we outlined in our 2012 TREE article, phylosymbiosis is a pattern that emerges under controlled conditions and will be subject to change with environmental dynamics.  Surely, bacterial OTUs can vary with environmental conditions and that variation is in fact one route to speciation by symbiosis that we raised in our TREE article in 2012. Different diets can unquestionably drive different microbiomes. The variation in microbiomes that were observed in our publications and ongoing research stems from variations in fly host microbes that the Nasonia are reared upon and the method of observing the microbiome (Evolution paper was cloning and sequencing, Science paper was Next-Gen). Furthermore, when we observe the bacterial sequences across our laboratory experiments and even in published genome sequences of Nasonia, we observe a conserved set of OTUs within the microbiome that spans nearly a decade of lab rearing; specifically Proteus, Providencia, and Acinetobacter.

In this current work and that of our 2012 Evolution paper, we independently demonstrate that when wasps are raised under the same environmental conditions, phylosymbiosis emerges.  But the phylosymbiotic microbiome itself does not have to be identical between environments, just as gene expression or epigenetics does not have to be identical between generations or environments. The remarkable thing in Nasonia is precisely that the UniFrac tree analysis finds phylosymbiosis independently in two published studies at different times and one more unpublished study in our lab. Thus, the phylosymbiosis pattern itself is stable, but the specific OTUs present in Nasonia do not have to be. We see this clarification not as a problem  but as a reflection of the natural way by which the host genome interacts within the microbiome.

We also are the first to admit that like any laboratory study of speciation, the processes observed in the laboratory are isolated observations whose importance is not whether they occur in nature, but that nature has the potential to operate and drive speciation events with them.

3. "Under their hypothesis, restoring the hybrid community to that of the parents (both of which have similar gut bacteria) should reduce lethality.  However, when the authors perform this experiment, they do not find significantly increased hybrid viability (Figure S1B; crosses g/v and v/g)"

We understand your position here. It would seemingly make sense that the resident microbial community from one of the parents could rescue hybrid inviability because they now have a normal microbiome.  But let's look at the data and the reasoning behind the experiments. First, Providencia is only one of the many species that naturally occur in Nasonia parentals as well as conventionally-reared hybrids. Thus, the mono-inoculation experiment here is not testing the specific prediction that a full, multi-faceted microbiome can rescue the hybrids. We previously tried this experiment by enriching for bacterial communities from parentals, but ran into several confounding problems of host material being present with the enriched microbial communities.  

Second, we're now skeptical that this prediction will actually hold water. Consider the gene-gene analogy from a speciation genetic perspective. You can only rescue hybrid mortality when you knock-out the epistatic genes, which is what we analagously did with the microbiome. But placing the "correct" gene in a hybrid background would only reinstate the epistasis that underscores the hybrid incompatibility. Thus placing the correct microbe, either specifically or a close relative, should recapitulate the epistatic interaction and mortality. Under this reasoning, your claim that a normal microbiome should rescue lethality is not supported either in our hologenomic view or a nuclear-centric view of hybrid incompatibilities 

4. " When the authors perform this experiment, they do not find a significant difference between Nasonia inoculated with their normal bacterial community (Providencia) or with a completely novel bacteria (Enterococcus)"

Some background on these two bacteria might be helpful here. Enterococcus is actually present in both hybrids and non-hybrids (we state that in the third page of the main text, also see our supplemental tables) and thus we expected it to kill because the host genome would recognize it. Ongoing experiments with very distantly related bacterial species that Nasonia may not be able to recognize do not seem to kill the hybrids, which is what we would predict. Thus, we assert that your reasoning here is backwards.

In summary, a major part of our effort is to convince evolutionary biologists that the microbiome arises from an element of host control and under host control they are part of an extended phenotype with genomes themselves. Microbial ecologists have weighed this evidence to be quite self-evident in the past few years, though not every study bears out an element of host control. Oddly enough, both fields accept the intimate interactions between the host genome and ancient endosymbionts / organelles; but clearly they emerged from predecessors that were more free-living and “farmed” from the environment until they became obligate and germline transmitted. What we are seeing here is a continuum of symbiotic interactions.

No comments:

Post a Comment