By the end of this practical you will be able to:
In the previous practical you ended up with a group of Arabidopsis thaliana genes that respond to a bacterial infection, and a suggestion of what those genes are doing. That is a real result. It is also a result about a small roadside weed that nobody eats and nobody grows.
Almost everything we know about how plants defend themselves was worked out in Arabidopsis, and there is a good reason for that. Compare it with bread wheat, one of the crops the answer is eventually meant for.
| Arabidopsis thaliana | Bread wheat | |
|---|---|---|
| Genome size | 135 million bases | 16,000 million bases |
| Sets of chromosomes | 2 | 6 |
| Protein-coding genes | about 27,000 | about 107,000 |
| Seed to seed | 6 to 8 weeks | 5 to 8 months |
| Space for 1,000 plants | a bench | a field |
| Adding or removing a gene | routine | slow and often variety-specific |
That is why the experiments happen in Arabidopsis. A PhD student can run six generations in the time a wheat breeder runs one, in a growth cabinet instead of a field. The catch is obvious: the finding is now in the wrong plant. Getting it into the right one means finding, in wheat, the gene that corresponds to the Arabidopsis gene you care about. That gene is called an orthologue, and this practical is about what that word means, how you find one, and how far you can trust it.
The 500 genes from the previous two practicals, looked up in Ensembl Plants:
Two genes in two different genomes can look almost identical and still not be the pair you want. Whether they are depends entirely on how the two copies came to exist, and there are only two ways that happens.
A species splits in two. One ancestral population becomes two separate species. Every gene the ancestor had is now sitting in both of them, one copy each. The two copies start out identical and slowly drift apart as the two species go their own way. A pair of genes related like this is called a pair of orthologues.
A gene gets copied inside one genome. Copying accidents are common in plants: a stretch of DNA gets duplicated, or a chromosome, or occasionally the entire genome at once. Afterwards that one genome carries two copies of the same gene. A pair of genes related like this is called a pair of paralogues.
Both kinds of pair descend from one ancestral gene, so both will look similar and both will turn up if you simply search for lookalike sequences. Only one of the two is the pair you want when you are moving a result from Arabidopsis into a crop. The diagram below lets you build the history yourself and see what comes out.
Pick two genes, put a finger on each, and trace both fingers upwards until they meet. The question is only ever what kind of event you land on when they do.
Nobody has written down a list of which wheat gene corresponds to which Arabidopsis gene. There are hundreds of thousands of genes involved and the two plants last shared an ancestor around 150 million years ago. The correspondence has to be worked out from the sequences, and there are two ways of doing it.
Best match in both directions. Take the protein your Arabidopsis gene codes for and search all wheat proteins for the one that looks most like it. Then turn around: take that wheat protein and search all Arabidopsis proteins. If you land back on the gene you started from, the two are each other's best match and you call them orthologues. This is quick and it is what "reciprocal best hit" means. It also assumes there is exactly one best answer on each side, which step 2 should make you suspicious of.
Build the family tree. Collect every similar sequence from every species you have, build a tree of how they are related, and then go through the tree branch point by branch point deciding whether each one was a species split or a copying event. That is the same exercise you just did by hand in step 2, done automatically for tens of thousands of gene families. It is slower and it is what Ensembl Compara does, and it is where the one-to-one, one-to-many and many-to-many labels come from.
Everything below this point is that second method's real output. Pick one of the 500 genes from the previous practicals and see what Ensembl has for it.
Steps 4 and 5 follow this choice, so you can come back and change it at any time. If you picked a gene you liked in the GO enrichment practical, use that one.
Identity is the percentage of the Arabidopsis protein that is exactly the same in the other species. High confidence is Ensembl's own flag for the orthologues its tree supports most strongly.
The table in step 3 is a list, and a list hides the thing that explains it. These ten species are not ten independent samples: they are the tips of a tree, some of them close cousins and some of them separated by hundreds of millions of years. Putting the counts back on the tree usually makes the pattern obvious.
Each branch point below is a species split. The number beside each species is how many orthologues of your selected gene Ensembl finds there, and the colour is the relationship it reports. A species with nothing at all is worth as much of your attention as one with six.
Two genes from the same dataset that behave very differently. Load one, look at the tree, then load the other:
Ensembl says these two genes are orthologues. That claim rests entirely on their sequences, so it is worth putting them side by side and forming your own opinion instead of taking the label on trust.
An alignment writes the two proteins one above the other and slides them along until as many positions as possible line up, inserting gaps where one of the two has gained or lost a stretch. Each letter is one amino acid, the building blocks a protein is made of. The alignment below is computed in your browser when you press the button, using the same kind of algorithm the databases use.
This uses the gene you picked in step 3. For each species the page carries the closest of its orthologues, which is the one worth looking at; the rest are still listed with their identities in the table above. Sequences are available for six of the ten species.
The identity figure counts identical positions as a share of the Arabidopsis protein, which is how Ensembl counts it too, so it matches the number in the table in step 3.
Everything so far ran on data packaged into this page. Ensembl Plants is where that data came from and where you would go in practice, so open the real pages for the gene you picked. These three links are built from your selection in step 3 and go straight there.
One last thing, and it is the thing that most often trips people up. Orthologue means shared ancestry, not shared job. The definition is entirely about history: these two genes descend from one gene in one ancestor. That is all it claims. It is a good bet that they still do something similar, because that is usually how it works out, and it is only a bet.
Genes drift into new roles. A gene may be switched on in the root in one species and in the leaf in another, or matter enormously to a plant that lives in the cold and not at all to one that does not. When a gene sits in a family of near-identical copies, as most of the ones you just looked at do, the copies often divide the original job between them, so no single one of them is the equivalent of your Arabidopsis gene. The orthologue is where you start looking in the crop. Whether it does the same thing is a question you answer with an experiment.