A couple weeks ago a paper came out with some pretty cool results and suggestions. The authors found that specific light sensing rhodopsins normally found in the eyes were also found in the mouthparts of the fruit fly, Drosophila melanogaster. Proteins being found in cells or organs outside of their normal functional location is not that unusual. The thing is, these types of rhodopsins are well known and essential proteins needed in the eyes of most known animals for sensing light, and are not known to function in other senses or locations. But in Drosophila these specific light-sensing rhodopsins also function as taste sensors – the authors showed that the rhodopsins were essential for the fly to detect very low concentrations of bitter (toxic) compounds.
Here’s the paper by Leung et al:
The geewizery part is their suggestion that opsins, the class of proteins which includes rhodopsins, evolved from chemical (i.e., taste) sensors.
Since our lab works with fission yeast, I thought I’d do some poking around. I had plenty of COVID-time augmented with some OCD.
I ran an online program called DELTA-BLAST comparing the amino acid sequence of one of the human rhodopsin proteins against the all the proteins in the Schizosaccharomyces pombe database – trying to find the best match(es). DELTA-BLAST is a program that searches for what are called “remote protein homologs” – proteins that share similar structure and function but in the course of long evolutionary time have lost sequence similarity. If you are into the math and coding of these kinds of tools (I’m not and you’ll see the pickle this leads me into), here is the paper describing this technology:
When I ran DELTA-BLAST on the full amino acid sequence of a human opsin, the single top match was a pombe protein called Map3.
Map3 is a GPCR that detects a pheromone mating factor necessary for the yeast to begin the meiotic cell cycle. It is a chemical sensor. Very cool huh? But wait a minute, back up a tad… Pheromones? Mating? Yeast do that? Yes, as it turns out, yeast do mate. Under normal conditions when there is plenty of nutrition, yeast grow and multiply happily by fission (the yeast cells grow longer over a couple hours and then split evenly into two roughly equal sized daughter cells). All the while they remain sterile and do not mate. But under starvation conditions, the genes for mating become activated and different cells express different mating types (called P and M). These opposite mating types each secrete different pheromones, P-factor or M-factor. And they also make Map3, the receptor which detects the pheromone M-factor. (The other pair is Mam2 receptor which detects P-factor). The binding of pheromone to detector is essential for mating to occur. Successful mating results in a stereotypical bent ascus which contains four spores each of which will grow into a regular yeast cell when the environment allows.
Rhodopsins, as we learned in the previous post, are GPCRs like Map3, and the paper by Leung et al suggested rhodopsins (or the broader class of opsins) may have evolved from chemical sensors (which Map3 is). So… is Map3 related to the type of chemical sensor that eventually evolved into opsins – or is this BLAST result a statistical fluke, a false positive?
To see whether Map3 was a one-off fluke, I BLASTED several other human rhodopsins and other members of the opsin class of proteins and they all returned Map3 as the best-ranking match. Perhaps if the various human opsin proteins are closely related, then if one matches best to pombe Map3, they all should as well. So a false positive result for one can easily return the same false positive for the others. No surprise.
But there’s another thing to consider. Rhodopsins are GPCRs, a huge class of proteins found in all known eukaryotic cells. Humans have almost a thousand GPCRs. Yeast have less than
a dozen. DELTA-BLAST returned a yeast GPCR as a best match to several human rhodopsins. That is not so surprising, is it? Perhaps it would be more surprising if more divergent opsins also matched to Map3.
I then tried the three well-known fruit fly rhodopsins, Rh1, Rh2, and Rh3… and they all also returned Map3 as the top high-quality match. Again, perhaps it is not so surprising that if fruit fly rhodopsins match best to pombe Map3, over a dozen human opsins would all do the same. The only human protein I tested that did not match strongly to Map3 was called RPE-retinal G protein coupled receptor.
By the way, if you want to try this yourself, here is the protein BLAST tool:
And the NCBI database containing all the protein sequences I tested is here:
How about if I BLASTed what must certainly be a very divergent opsin, one called encephalopsin from an urchin (Hemicentrotus pulcherrimus)? Map3 was the best match of only two strong matches that DELTA-BLAST returned. It turns out that sea urchins and sea stars express opsins in specialized organs known as tube feet, and echinoderms (the phylum to which these organisms belong) lack eyes yet are reported to be light-sensitive.
This BLASTing began to fall apart in a different protein in a simple microscopic round worm (Caenorabditis elegans), where a UV light-sensitive protein called LITE1 did not match to pombe Map3. C. elegans have no eyes and are not known to be light sensitive. This result is interesting as well. Here we have a worm protein, LITE1, which is very closely related to 7-transmembrane taste receptors (a chemical sensor) and which was shown to enable sensing of UV light so the worm can avoid radiation damage. However, according to DELTA-BLAST, this LITE1 protein does not match at all to pombe Map3, a chemical sensor.
But BLASTing opsins from other worms, such as the more complex segmented worm Platynereis dumerilii which have the simplest eyes known in the animal kingdom, returns a single strong match to Map3 and no other pombe protein.
I tried opsins from various simpler organisms like planaria and in some I got strong matches to pombe Map3 (Schmidtea mediterranea, Schistosoma japonicum, Sagartia elegans) or poor or no matches (Helobdella robusta).
Then I went all the way down to opsins from various single-celled organisms… and in most cases Map3 was a weak match or no match at all. I also started to get pombe Mug73 as a top return for some of the rhodopsins in organisms such as Chlamydomonas reinhardtii, Salpingoeca rosetta, Halobacterium salinarum, and Allomyces arbusculus.
What is Mug73? According to pombase which is an online S. pombe database, Mug73 is a “multispanning 7TM plasma membrane rhodopsin family protein, implicated in signaling” and it has protein features that match best to archaeal/bacterial/fungal rhodopsins. Mug73 does not appear related to Map3 and BLASTing does not indicate any close match between these two 7-transmembrane proteins.
But ain’t this interesting? Rhodopsins from humans (and from many other divergent species) match strongly to a pombe Map3 pheromone sensor which is not classified as a fungal rhodopsin, and conversely does not match at all to Mug73 which is classed as a rhodopsin family protein.
At this point I am like a lab rat in his COVID cage, pressing the food bar and getting his pellet. The rat doesn’t understand the mechanics behind how pressing the bar gets him his reward… just that pressing the bar gets him what he wants.
What if we reverse the BLASTing? What if I take the pombe Map3 and DELTA-BLAST it against the human protein database? I press the food bar… then I get back many strong matches to proteins such as somatostatin receptors and many other receptors. But I also find that human melanopsins, Opsin 4, Opsin 5 and others also show up as strong matches to Map3 though not the top ones. Reward!
If I take pombe Mug73 and DELTA-BLAST it against the human protein database, nothing comes back as a match.
There are several pombe GPCRs besides Map3, and these include Mam2, Git3, Ste3, Stm1… and when I BLAST these against the human protein database various proteins come back, but none match to any opsins. Press the bar… press the bar…
Finally, I decided to use the Clustal-Omega tool that I used in a previous post to make an evolutionary tree of Cdc25 from divergent species. I thought to try that same tool on various opsins across a similarly broad range of species.
This is what came back.
We can see several human opsins clustering very close together and having very short “twigs” such as the melanopsin isoforms near the top and the human opsins at the very bottom. The interesting thing is that this program assigns pombe Map3 to a cluster closest to LITE1 from C. elegans (also known as High-energy light unresponsive protein 1), followed by human Opsin 5, and then the other human opsins including Opsin 4 (Opsins 4 and 5 came back as strong matches to pombe Map3 when I did the reverse BLAST exercise).
The 1’s and 0’s I appended to the end of each line denote if DELTA-BLAST returned Map3 as a strong match to it (1) or not at all (0).
I made a similar map with mostly proteins that did not match to Map3 and this is what that tree looks like:
All the proteins that BLASTED best to Mug73 all cluster together which I think is a good validation of this clustering program relative to the DELTA-BLAST results. Again, in this group of proteins mostly not related to Map3 (other than human Opsin 1), pombe Map3 clusters better to C. elegans LITE1, and is only distantly related to human Opsin 1. And the two human opsins cluster together – again a good validation.
What if I combined both lists – proteins that do and do not match to Map3? What would Clustal Omega spit back at me?
Again, the 0’s at the end of each line means that DELTA-BLAST program did NOT match that protein with pombe Map3. The 1’s at the end of the other lines indicates that DELTA-BLAST returned pombe Map3 as the best high-scoring match to that protein.
Here it is.
All of the 0’s EXCEPT pombe Map3 all cluster together. Map3 sits in the middle of all those proteins that DELTA-BLAST says are not matches – like a sore thumb or a different digit.
All the proteins boxed in red in the previous tree – that all DELTA-BLASTED most closely to pombe Mug73 – again all cluster closest together in this bigger tree. Again, a good validation.
The drosophila opsins all cluster together. As do most of the human opsins.
And yet again pombe Map3 clusters most closely with C. elegans LITE1 protein… and remember LITE1 is a protein that is closely related to taste receptors (a chemical sensor) which on one hand makes sense since Map3 is a pheromone receptor (a chemical sensor). And LITE1 has been shown to function as a UV sensor for the worm, which has no eyes.
Is this clustering real – as in accurate?
Or is this clustering of Map3 with Lite1, and in the middle of non-matching proteins, an artifact or limitation of software, glitches or deficiencies in either DELTA-BLAST or in Clustal Omega? I think that is always a possibility especially in programs I view as black boxes. I put in numerous sets of proteins that are closely or somewhat related, and the program seemed to detect and cluster those well. But it is one thing to cluster proteins that are similar – I imagine it is another thing altogether to detect or correctly order relationships among proteins that are more divergent – more different. Have I gone beyond the limits of the DELTA-BLAST and Clustal Omega software’s ability to detect remote protein homologs?
If we look at Clustal Omega’s alignments of select regions in the protein sequences of all the proteins in that tree, we see the color-coded image below. The region I picked is one that I thought looked like the strongest match among the most proteins. The proteins are not in the same order as in the phylogenetic tree.
The first two rows are proteins that are very different from all the others and it is no surprise that there is little matching (they are the same as the first two in the combined tree). The bottom seven rows are the proteins that match best to Mug73 – so that is also good. The 8th one up CAA93308 is Map3. Looking at the sequences, it is no surprise that Clustal Omega clustered Map3 with this batch of Mug73-like proteins. CCD64541 is the C. elegans LITE1 protein – which I think looks nothing like Map3 at this portion of the sequence. The remainder of the proteins in the middle rows that all match well here are the various opsins that BLASTed well to Map3.
Visual inspection of the rest of the sequences doesn’t improve the sequence matches from this snapshot here. There isn’t a part of the protein where suddenly pombe Map3 aligns (by eyer) better with the human opsins, for example.
I remain intrigued and puzzled by these protein alignment results.
Clustal Omega is clearly able to detect regions of homology, or similarity, between a wide range of opsin proteins across a vast gulf of evolutionary separation from single celled yeast to human. Cdc25 was a better example of that as seen in a previous post.
The puzzle is how Map3 was consistently returned by DELTA-BLAST as the top or even the only best match to various rhodopsins in various species, but in the Clustal Omega program Map3 is never clustered with those proteins but instead placed smack in the middle of all the proteins that DELTA-BLAST said were not related to Map3. And further, Clustal Omega consistently places Map3 closest to the C. elegans LITE1 protein which DELTA-BLAST denied any similarity to, though they both have a relationship to chemical sensor proteins.
In any case, I think the C. elegans LITE1 protein is a good example of a protein that spans the chemical and light sensing domains, as suggested in that paper by Leung et al that started this whole exercise in BLASTing.
And the question remains whether pombe Map3 is the type of chemical sensor that, in some common ancestral species and by some meandering evolutionary paths emanating from that remote branch point, yielded a light sensing precursor to our diverse family of opsins. Or if I am abusing the abilities of the software that remains a black box to me.
Comments