Smoking a herring turns it red and imbues the fishy flesh with an extra strong odor, which can be used to throw bloodhounds off your trail, or to train hunting dogs to stay on the main scent trail, even if far less odoriferous. Such are a few hypotheses for the origin of the term "red herring", used to denote an enticing, yet ultimately uninformative beacon.
I've come to realize that my own field of eye evolution has a resoundingly pungent red herring that continually throws scientists and non-scientists alike off the trail of a deep understanding of how eyes, and more generally, photoreception, evolved. In this case, the rank fish is the simple and seemingly harmless question, "How many times did eyes evolve?"
Quite often, when I tell a newly acquainted colleague that I study eye evolution, he or she will pull the red herring out of their back pocket, and dash across the room, dragging the fish behind them. "So how many times did eyes evolve, anyway?” they ask, hoping I have a key to this deep mystery. When I was a younger bloodhound, I often took that bait, howling and bounding enthusiastically after that trail, circling the room again and again. But I've followed that trail enough times by now to know that it circles back on itself in a confusing and illogical way.
How many times eyes evolved, though enticingly simple, is an all but unanswerable question; at least in terms of a numerical answer that people crave. Eyes did not evolve one time, nor did they evolve forty to sixty-five times. The answer is that some components of animal eyes are shared in all animal eyes we've ever examined. Yet some components are new comers to organs that probably functioned as eyes even before those components joined the party. Asking how many times eyes evolved makes the implicit assumption that every component of that eye shares a congruent history, that an eye is either all there, or all not there. But eyes are complicated structures, and the evolutionary histories of its parts are varied.
A new paper by Kozmik et al in PNAS is an interesting piece of science that breaks new ground. Although it does not detract from the scientific value of their study, the authors have been beguiled by our ripe red herring. Still, they have elegantly streamlined the malodorous query in a way that will allow me to more easily explain the stench, and so I retell their work here. First, I’ll explain some of the new data then I’ll point out where I think they, like many others before them, lost the main scent trail.
The authors examined a box jellyfish, a group that includes the infamous sea wasp, the deadly nemesis of Australian beach-goers. Box jellies are cnidarians, the group including corals, anemones, and hydras, which besides sponges, may be the most distant animal relatives of humans. As such, the biology that is shared between humans and cnidarians likely originated very early in the history of animals, hundreds of millions of years ago. With this logic in mind, Kozmik et al focused on elucidating some of the components of the eyes of one particular box jelly that lives near Puerto Rico, a far less noxious variety than the Australian sea wasp. In particular, they elucidated some photoreceptor genes and some genes that produce a shading pigment in the box jelly’s eyes. Choosing these two particular components was no accident. They were chosen because they represent a common minimum definition of “an eye”. Some scientists argue that minimally, an eye must contain photoreceptor cells to register the presence of photons, and some type of shielding pigment to allow the owner to determine the direction of light. By focusing on these two components, Kozmik and colleagues were streamlining the question of whether eyes evolved more than once, asking whether we could find evidence that photoreceptor plus pigment joined forces multiple separate times, or whether their partnership dates back to a singular union.
Simplifying the real biology a bit, there are two primary classes of phototransduction, which is the cellular process sensing a photon and firing a nervous signal. One pathway that I will call “r” was first elucidated in flies, the other that I will call “c” was first elucidated in vertebrates. It was once thought that the r-pathway is an invertebrate trait, and the c-pathway is a vertebrate trait. However, the last 5 years have brought good evidence that c and r are present in both vertebrates and invertebrates, although the c-pathway is a bit cryptic in invertebrates, and the r-pathway is a bit cryptic in vertebrates. Kozmik et al investigated three components of phototransduction, all three components showed some evidence of being more similar to the “c” class of phototransduction. This primary conclusion led Kozmik et al to choose their title “Assembly of the cnidarian camera-type eye from vertebrate-like components”.
The first component they studied was opsin, a protein that binds to a light reactive chemical to initiate light perception in animals. With opsin, they argued for “c” type phototransduction, as opposed to the “r” type of fly compound eyes. Opsins form multiple sub-families, and the box-jelly opsin might be within the c-opsin subfamily based on their phylogenetic analyses. However, the analysis presented in the supplement of this paper is actually rather inconclusive about c-type status, although it is definitely not an “r”-type. We did a quick reanalysis and found the box jelly opsin might actually belong to a cnidarian-specific opsin sub-family that we discovered last year and which we named “cnidops”. None of the cnidops genes were included in the Kozmik phylogenetic analysis.
Regardless of its sub-family type, this opsin was found to be expressed specifically in the box jelly eyes. One of the most interesting new pieces of science in this paper is that Kozmik et al were able to match the function of the protein itself with that of the jellyfish eye. To do this, they functionally activated box jelly opsin in cell culture and measured the protein’s function. Cell culture is the practice of growing isolated animal cells under controlled conditions, without an animal itself. Introducing the box jelly opsin protein to cells in culture allowed the protein to fold properly and bind to a light- reactive chemical that was separately introduced into the cells. By shining lights of particular wavelengths (“colors”) on to the cells, the authors could see which colors of light activated the box jelly opsin protein. As expected if the opsin in question is used in box jelly eyes, the authors found that the opsin is sensitive to similar wavelengths as the jellyfish eye itself (light we would call blue-green, of about 465 nm wavelength). The match between the animal's light response and that of the expressed opsin protein is a very good piece of evidence supporting the idea the opsin they studied is involved in the animal's light response. This is one of the first times that an invertebrate opsin has been studied in cell culture, and certainly the first Cnidarian opsin, so this is a very exciting experiment.
Evidence for other components of the “c”-pathway being used in box jelly eyes are preliminary. The authors did find that the mRNA of genes similar to vertebrate c-pathway genes were present in the box jelly eye. However, this was as far as the present study went with the non-opsin components. Namely, they did not yet investigate protein expression, nor did they yet conduct any biochemical experiments to demonstrate the interaction of opsin and the other phototransduction components they found. Granted, these additional experiments represent a lot of work. The authors also did not report any experiments to show that r-pathway or other similar genes are not expressed in box jelly eyes (but I note they did not mention the presence of these genes in their library of genes from the box jelly eye, so perhaps they really are absent from the library). In the end, it remains a bit of a leap of faith to conclude that the opsin is actually interacting with the other genes, reported only from mRNA. Nevertheless, the authors have taken some important steps in that direction, so for this essay, we can give the authors the benefit of the doubt that a c-pathway is interestingly being used in the box jellyfish eyes. Even if our hunch is true that it is actually our new “cnidops” class of phototransduction, it doesn’t change the main message here; the herring will remain red.
The second criterion, besides the presence of photoactive cells, required to satisfy a minimal definition of an eye, is the presence of one or more pigmented cells. Here again, Kozmik et al found that pigment-components of a box jelly eye are very similar to pigment-components of vertebrate eyes. Here, the data are quite conclusive. The authors found genes homologous to vertebrate melanin genes expressed in parts of the jellyfish eyes that contain abundant pigment. In addition, there is a direct chemical test for melanin, which box jelly eyes passed. Interestingly, box jelly cells have dual function as photoreceptor and pigment-bearing cell.
Based on these data, Kozmik and colleagues favor the conclusion of parallel evolution – the separate assembly of homologous (“the same”) components to form eyes. In other words, they argue that box jellies and vertebrates separately evolved eyes, but happened to use the same components in each case, opsin and melanin. This conclusion is a bit of a curveball given how I introduced the study above. The conclusion is that the jelly and vertebrate eyes share the same components, so why do they not conclude a common ancestry of the eyes? They actually do point out that this is an alternative possibility. However, they favor parallel evolution based on what boils down to inferring phylogenetic history of these components and of eyes more generally. Unfortunately, they only use verbal arguments, instead of the well established statistics of phylogenetics and ancestral state reconstruction, which could’ve been used to test their claims statistically (see this paper for examples).
The main verbal argument of Kozmik for parallelism, and against common ancestry, is that many animal phyla do not have eyes. As such, it would seem that eyes would have to be lost in many phyla, if common ancestry of box jelly and vertebrate eyes holds (this is where a real phylogenetic statistical test would be nice – undergraduate honor’s project anyone?). In addition, they point out that cnidarians and vertebrates utilize different transcription factor genes in the specification of eyes during development. So basically, they are assuming a priori, based on previous evidence, that the eyes are of “independent origin”, and so based on that assumption, it is surprising to find that they use the same components.
Back to the red herring
So, both vertebrates and box jelly eyes may use c-pathway photoreception and both use melanin as a shielding pigment. You might be wondering where I stand – for parallelism or for common ancestry? Arrrooofff. Bow, wow, wow. Did I catch you? Has your mind been bounding down that pungent scent trail, sniffing and howling enthusiastically? Or have you kept the main, fainter trail, taking my advice that this question of number of origins is misplaced? Perhaps some more explanation will help you understand my point, and wizen your inner bloodhound. Let’s follow the red herring trail and see where it leads.
First, in case it’s not already clear, I’ll point out that the question of parallelism versus common ancestry of box jelly and vertebrate eyes is asking about the number of eye origins. It is our red herring question. Parallelism implies two origins, two separate unions of photoreceptor and pigment. Common ancestry implies a single merger, at least for the two eyes in question.
The reason I don’t like the red herring question, is that it puts the focus on “the eye” rather than on the components that define the eye. The Kozmik et al study is about the components – they elucidated for the first time some of the molecular components of a box jelly eye. This is noteworthy. But what do we gain from saying that eyes originated more than once? I think very little, and in fact such a conclusion masks some of the interesting biology. For example, stating that “eyes originated more than once” fails to recognize the common ancestry of some of the components, notably opsin. Any animal eye that has been examined to date, in addition to photoreceptive cells not in eyes, utilizes a homologous gene family (opsin). This point gets lost when simply stating that eyes evolved more than once.
A related difficulty is that this perspective neglects an intermediate stage in the complexity of photoreceptive structures. Someone unfamiliar (say an anti-evolutionist) who reads “eyes evolved more than once”, imagines the genesis of complex eyes from nothing. But that is not how evolution works. Evolution uses building blocks used for other functions in new combinations. This is illustrated by the conclusion of parallelism of Kozmik et at. In their model, opsin-based phototransduction existed unassociated with pigment cells. Once phototransduction and pigments became associated, perhaps separately in the box jelly and vertebrate lineages, the organisms came to posses a minimum definition of “an eye”. But this definition is just a human construct. Again, to my mind it obscures the fact that much of “the eye” was already there, it just took a new association of components (photoreception plus pigment) to evolve an eye. [For the cynics arguing that this just pushes back the real origins, I recognize this, which is why my lab is studying the origins of phototransduction – by studying the evolutionary history of each of the components separately. The bottom line – phototransduction evolved by using existing and newly duplicated and diverged components, a common theme in evolution].
You may be asking what if the opposite answer of the red herring question holds. What if we conclude “eyes share a common ancestry”, which could be concluded based on the Kozmik et al data of shared opsin and pigment components in vertebrates and box jellies. Again, I would argue that we gain little by making this statement. In this case, such a statement can only be made for the components in question. Perhaps it can serve as a sort of “null hypothesis”, where it serves as a prediction for new data that we don’t yet know about, which has some value. However, we already know that for eyes, not all components have the same history, which is implied by the statement “eyes only evolved once”. We already know that different lens proteins are used in different types of eyes. We also already know that components of phototransduction cascades differ. For example c-phototransduction utilizes the ion channel protein CNG, whereas r-pathways utilize the unrelated ion channel protein TRP. Saying that all eyes have a common ancestry does not account for these details. Again, it is about the components. Different components have different histories, and this cannot be accounted for in the simple dichotomy of “one versus more than one origin”.
The question of the number of eye origins has been around for a while now. But I argue that it is time to move beyond this question, and to recognize it for the red herring that it is. The fact of the matter is that some eye components are shared and others are not, and this varies depending on the time scale and species examined. Of value is to fill in the narrative of this amazing evolutionary story. We now have the ability to understand to the level of specific molecular changes how vastly different types of eyes evolved. This is the main – though perhaps fainter – scent trail.