Endosymbiosis: the feeling is not mutual

Abstract

Endosymbiosis was an idea that provided a remarkable amount of explanatory power to observations about eukaryotic organelles. But it also promoted a few assumptions that have been less well-examined, and here we look at two of these. The first is the idea that some endosymbiotic relationships that are assumed to be mutualistic, such as nutritional symbioses and eukaryotic organelles, are not in fact power struggles mush as we assume many other ecological interactions to be. The second is that endosymbiotic merger between organelles and their hosts involved the acquisition of a great many genes that took on functions in the host. New data from other endosymbiotic systems and the organelles themlseves suggest some of our hypotheses about organelle origins and distribution may be misled by the expectation that such genes exist and persist in large numbers.

Introduction: Untangling what we know and what we assume about endosymbiosis

The idea is old and appealingly simple: dissimilar organisms live together and by doing so become more than they were as individuals (De Bary, 1879; Sagan, 1967). They become a symbiosis. It is clear that symbioses have propelled associations of organisms into environments where the individuals alone could not survive, and by doing so have massively affected the evolution of life (Archibald, 2014). But what is less clear is how entering into symbiosis affects the participating organisms. Symbioses are often described as mutualisms, or relationships where both partners benefit. (In fact, the word symbiosis itself is sometimes used interchangeably with mutualism.) But the benefits for both partners are sometimes hard to see. Indeed, mutualisms have been previously likened to symbioses that have simply found a way to manage the inherent conflicts of interests between organisms (Herre et al., 1999). This is especially true in endosymbioses, which can often look remarkably one-sided (Bennett and Moran, 2015; Garcia and Gerardo, 2014; Kiers and West, 2016).

In this paper we use symbiosis in its most general sense. That is, we define it simply as any sustained organismal interaction somewhere on the pathogenic-beneficial continuum (Lewis, 1985). We highlight several recent examples that expose how the evolutionary interests of endosymbionts can become misaligned, and how endosymbioses that seem extremely interdependent and stable (even “permanent”) can break down under the right circumstances. In particular, we focus on two aspects of endosymbiosis that affect our thinking of evolution more broadly: the idea that endosymbiosis is often a mutualistic relationship, and the idea that endosymbiosis has a deep and lasting an impact on the genome evolution through endosymbiotic gene transfer.

Endosymbiosis as an antagonistic relationship: Context matters

Putting aside the classic organelles, the mitochondria and plastids, the vast majority of endosymbioses are probably pathogenic. That is, the presence of a microbe inside a host cell most often imposes a cost from the host perspective. If this is true, it follows that most endosymbioses that are beneficial from the host perspective likely evolved from interactions that were initially pathogenic, or at least mildly so. This idea is supported by analyses of the origins of proteobacterial symbionts with various hosts, which shows that the vast majority of beneficial proteobacterial symbionts have evolved from pathogenic ancestors (Sachs et al., 2014). What is required for an endosymbiosis to shift from costly to beneficial from the host perspective? Quite simply, the ecological context must change so that the benefits of the interaction outweigh the costs. The ecological context shifts that seem most common in endosymbiosis are those involving hosts gaining access to previously inaccessible nutrition or energy, or those where hosts defend themselves in ways not possible without the presence of the symbiont.

In endosymbioses where the microorganism provides energy or nutrition for the host, such as the mitochondrion, plastid, and many nutritional symbionts in insects, the context shift is absolute and (nearly, seemingly) permanent: the host cannot survive in the environment without its symbiont. These sorts of massive ecological context shifts drive the most spectacular and long-term types of symbiosis, because the host must preserve the symbiont at all costs (or get a new endosymbiont, acquire the function in some other way, or move to a new environment). The long-term and strictly dependent nature of these symbioses can make the context dependency hard to see, because loss of endosymbiont without a change in host context results in extinction of the entire symbiosis. However, the context dependency of symbiosis is often clear in symbioses that are relatively recent associations, such as protists and their photosynthetic symbionts (Lowe et al., 2016) or amoebae and their bacterial symbionts (DiSalvo et al., 2015). But recent work provides a few instances where long-term endosymbioses—the type perhaps more naturally thought of as mutualisms—seem to be in the process of breaking down or have actually proceeded to eliminate their endosymbiont.

We can gain some insight into the way we instinctively think about endosymbiosis by considering the case of an insect endosymbiont called Hodgkinia cicadicola. Hodgkinia is in many ways a typical insect nutritional endosymbiont. It provides cicadas with two of the ten amino acids that they cannot make on their own and that are not provided at high levels in the strict plant sap diet of the insect (the remaining 8 essential amino acids are provided by another bacterial endosymbiont called Sulcia; McCutcheon et al., 2009). In many cicada species, this clean narrative is preserved: Hodgkinia provides two essential amino acids, Sulica provides the other eight, and the host gives them a nice place to live. Everyone is happy, and from some perspectives it looks like a three-way mutualism.

But this tidy story starts to break down in other cicada species. In some cicadas, the single ancestral Hodgkinia lineage has fragmented into two new distinct cell types, each with a distinct genome that has lost genes so that both are required by the host to provide the nutrition required by the ancestral single lineage (Van Leuven et al., 2014). Put another way, a host that used to have to keep track of two bacterial lineages (Sulcia and a single Hodgkinia lineage) now is required to keep track of three. Why does this happen? It isn’t clear yet, but we suspect that it is related to the unusual, long, and variable life cycles of cicadas. Documented cicada life cycles are between 2 and 17 years, and we know that in a short-lived species there is one Hodgkinia lineage, and that in the longest-lived cicada species there are several dozen Hodgkinia lineages (Campbell et al., 2015). These long-lived cicadas must therefore cope with numerous Hodgkinia lineages, each one encoding just a few genes.

While vertical transmission normally promotes cooperation between host and symbiont (Bull et al., 1991; Ewald, 1987), this cooperation seems to be breaking down in some cicada groups despite an unchanged vertical transmission route. What has changed? We suspect that the high mutation rate of Hodgkinia combined with the increased symbiont generations that become possible in long-lived cicadas increases the genetic diversity of the symbiont population and thus promotes competition (or selfishness) and limits cooperation. Less fit symbiont genotypes can rise to high frequency during long cicada generations and occasionally get fixed, in an event we see as a split lineage (Van Leuven et al., 2014). From the host perspective, this process is probably nonadaptive. It is not better to transmit dozens of Hodgkinia lineages to each egg instead of one, but the host has little recourse (outside of symbiont replacement) because its ecological context requires the two amino acids that Hodgkinia still produces. The cicada is stuck in a symbiont rabbit hole of its own making (Bennett and Moran, 2015).

What makes Hodgkinia different from a classical intracellular parasite such as the malaria parasite, Plasmodium? We argue it is primarily context—or, different only in the direction in which the hostility is aimed. For Plasmodium, the simple narrative is the infectious agent is forcing itself into a cell, disrupting its normal function and subverting it to its own purpose. Ultimately the infectious agent kills and discards its host to move on to take over the next hapless victim. In the case of Hodgkinia, the story initially seems more like a nurturing embrace of one cell by another, the bacterium enveloped by its host, allowing it to shed functions that are no longer needed because essential nutrients and energy are all readily donated by its new benefactor. But when the context changes—in this case, the host life cycle—the cooperation between host and symbiont breaks down. The selfish tendencies of Hodgkinia were always there, they were just held in check by the host.

We suggest that endosymbiotic interactions are best thought of not as mutualistic “happily ever-after” stories, but instead as “use it up and cast it off” situations that are stable for variable lengths of time. Endosymbiosis nearly always produces dead ends for one of the two partners—in the case of Plasmodium and other traditional parasites, the host is the partner that is cast off in the short term, but in the case of Hodgkinia and other beneficial endosymbionts it is the symbiont that is cast off in the longer term. In one case the symbiont is exploiting the host, while in the other the host is exploiting the symbiont. But neither one is mutualistic: they are both power relationships that differ simply based on whether the internal or external partner is in control.

This logic suggests that endosymbiotic relationships will always be temporary and they will be lost or replaced, but don’t we already know this is not true? The answer depends on the time scale one considers, and the amount of diversity one has studied. For example, a taxonomically narrow view of the bacteria-in-bacteria mealybug symbiosis might lead one to conclude that this baroque structure has only evolved once, but a view with a wider taxonomic breadth and depth reveals frequent endosymbiont turnover (Husnik and McCutcheon, 2016). But what about eukaryotic organelles? Are they not the classic case of endosymbiosis leading to “happily ever after”? It’s clear that the host has control, so why has the endosymbiont not burned itself out like Hodgkinia seems to be doing? Why have organelles not been replaced with fresh symbionts? It could be that organelle degeneration has stabilized due to large amounts of gene transfer and protein-targeting. But it is also clear that if the core function of the organelles is acquired independently or side-stepped somehow, that even this “permanent” relationship can be lost. Indeed, photosynthesis has been lost scores of times in plastids and oxidative phosphorylation and electron transport has been lost many times in mitochondria when the host’s ecological context has changed such that these functions were no longer required (Burki, 2016; Müller et al., 2012; van der Giezen, 2009; Williams and Keeling, 2003). As highly valuable as these functions are, they are not core functions of these organelles. Instead, the core function of mitochondria, the last function to be retained in even the most reduced organelle, is iron-sulfur cluster assembly (Müller et al., 2012; van der Giezen, 2009; Williams and Keeling, 2003). In the case of plastids, the core function is likely different in different lineages, but fatty acid, amino acid, heme, and isoprenoid biosynthesis are all candidates in different groups which have lost photosynthesis (Foth and McFadden, 2003; Williams and Keeling 2003; Keeling, 2013; Ralph et al., 2001). But recent studies that have considered eukaryotic diversity more broadly show that even these core functions have sometimes been lost, and when they are, the association breaks down and the endosymbiont is eliminated (Gornik et al., 2015; Janouškovec et al., 2015; Karnkowska et al., 2016). Moreover, in the dinoflagellates, where photosynthesis and even plastids are particularly prone to loss, we see occasional cases of plastid replacement: apparently when one plastid is used up and discarded, another one can be acquired to replace it (Keeling, 2013; Archibald, 2015). These studies show that even organelles have not been frozen into permanence, it just seems that way because we have not looked broadly enough (or waited long enough). The degenerative ratchet is still slowly turning, even in organelles.

Overall, when we mentally distinguish “cooperation” and “competition”, it is a mistake to apply the kind of thinking that we intuitively glean from the relationship within macroorganism symbioses such as cleaner wrasses and fish to the rather more abstract relationship we observe between one cell living within another (although they may be more similar after all, since context-dependent breakdown of the wrasse-fish mutualism is observed: Gingins et al., 2013). Instead, we argue that endosymbioses are rarely, if ever, mutualistic. Endosymbioses are just different forms of competition, where the vector of control points in different directions with different magnitudes depending on the context. One partner is always in control, or fighting to increase control. Coexistence can occur for long periods of time, but if conditions change the partnership can quickly tip towards extinction for either the subordinate member or the entire symbiosis. The interesting questions for long term endosymbiosis, like eukaryotic organelles, therefore shift from why and how the partnership formed, to why and how the partnership has so far avoided extinction.

Genomic impacts of endosymbiosis

The endosymbiotic origin of mitochondria and plastids primed us to accept similar explanations for other phenomena. At the cellular level, this initially led to a rush to explain other organelles in endosymbiotic terms, for example the flagella and cilia, peroxisomes, endoplasmic reticulum, and even the nucleus (Cavalier-Smith, 1987; Gupta, 1999; Lake and Rivera, 1994; Margulis, 1970; Sagan, 1967). Endosymbiotic explanations for these organelles has gone out of fashion due to an ongoing absence of evidence (Keeling, 2014; Martin, 1999), but the fashion has made a comeback to explain genomic data. The acceptance that mitochondria and plastids were indeed derived from endosymbiotic bacteria came at an auspicious time in the early days of molecular biology and subsequently genomics. These technologies were revolutionary, broke down a lot of long held ideas, and lead to an intellectual vacuum to be filled with new explanations. At least some of this vacuum was naturally filled by explanations involving endosymbiosis (Keeling, 2014); some of these explanations have now formed the foundations for other assumptions, but have not been subjected to serious critical examination.

Most important of these is a prevalent idea that by looking at the evolutionary history of genes in a genome we can “see” an ancient endosymbiosis based on the presence of genes in the host that were acquired from that endosymbiont. This presupposes that an endosymbiont will donate genes to the nucleus of its host, an idea with a complex history. Almost simultaneously with Margulis’ influential paper in The Journal of Theoretical Biology (Sagan, 1967), Goksøyr outlined a similar hypothesis in Nature (Goksøyr, 1967), and went further suggesting that the endosymbiont would have moved some of its genes to the host, and that their products would then be targeted back to the endosymbiont. Weeden (1981) developed this idea further, going as far as to say it was a necessary corollary to the endosymbiont hypothesis because organelle genomes were insufficient to encode the necessary genes to support organelle function. Although the exact origin of all nucleus-encoded organelle genes has been questioned (Keeling, 2013; Larkum et al., 2007), these two ideas have provided enormous explaining power when looking at organelle biology and evolution. However, beyond this well-tested core is a less-well-examined idea that is nevertheless influential. The thinking goes that if a large number of genes were transferred to the host for proteins now targeted back to the organelle, then probably a lot of other genes were transferred as well. Many of these proteins, if not most, are now not targeted back to the organelle but acquired functions in the host.

This seems reasonable enough - genes were flowing, and if they are potentially useful then it stands to reason the host should keep some to function in cytosolic pathways, and maybe even keep a lot. Early studies supported this conclusion based on genomic data from model systems (e.g., Martin et al., 1998). Naturally, the implications of this conclusion can be extrapolated to touch on other, more complex problems. Most importantly, if organelle endosymbionts donated a lot of genes for now-cytosolic proteins, then we should be able to “see” evidence for now-lost organelles in the nuclear genomes of their erstwhile hosts. This idea rests on the assumption that, because these genes have acquired a function independent of the organelle, they will be retained even when the organelle is lost or replaced.

If true, this would be a powerful tool in the reconstruction of evolutionary history, and has formed the logical basis for a number of claims for ancient endosymbiotic events and cryptic or now-lost organelles. For example, in work on plastid organelles, such studies have concluded that non-photosynthetic lineages like oomycetes or ciliates once had a red algal plastid (Reyes-Prieto et al., 2008; Tyler et al., 2006), or that red algal plastid-containing lineages once had green algal plastids (Moustafa et al., 2009; Woehle et al., 2011). These conclusions have been challenged on the basis of the veracity of the phylogenetic results (Burki et al., 2012; Deschamps and Moreira, 2012; Moreira and Deschamps, 2014), but the idea itself has not been challenged particularly, and has had a major impact on models for the evolution of organelles and on how we perceive the impact of endosymbiosis on the host genome and cellular function. Indeed, it emphasizes the importance of endosymbiosis on both counts: it predicts more endosymbiotic organelles in evolution and ascribes more functional impact to them. However, these conclusions are dependent on an assumption (that organelle derived genes will be kept in large numbers when the organelle is lost) that is itself built on another assumption (that those genes were transferred and retained in the first place), and neither has been thoroughly tested. In the almost two decades since the original analyses supporting the presence of large-scale transfers of genes from the organelle endosymbiont for proteins that do not function in the organelle (e.g., Martin et al., 1998), there have been significant advances that would allow this important conclusion to be reexamined with more confidence. We are now awash with recent genomic data from a variety of eukaryotes, and phylogenetic methods and computational power both now allow for significantly better tests of a gene’s origin. Some studies support an overall episodic influx of genes that is consistent with this idea (e.g., Ku et al., 2015), but other studies on relatively recent secondary and tertiary endosymbiotic events find the number of endosymbiont-derived genes in the host nucleus that that are not functionally linked to the organelle to be few, or even potentially zero (Burki et al., 2012; Curtis et al., 2012; Hehenberger et al., 2016; Moreira and Deschamps, 2014; Patron et al., 2006). Different endosymbiotic events may have had different impacts, but if the assumption is untrue, or even if significant variation is found in different organelle origins, then it will limit the extent that we can interpret the presence or absence of such genes from a nuclear genome. This in turn impacts how much weight can be given to endosymbiosis to explain eukaryotic diversity.