Wednesday, March 19, 2008

Did Pre Mammalian Therapsids Have Milk?

A new paper by David Brawand, Walter Wahli, and Henrik Kaessmann investigates the transition in offspring nutrition by comparing the genes of representatives of these three different mammalian lineages with those of the chicken—an egg-laying, milkless control. The authors found that there are similar genetic regions in all three mammalian lineages, suggesting that the genes for casein (a protein found in milk) arose in the mammalian common ancestor between 200 and 310 million years ago, prior to the evolution of the placenta.

Eggs contain a protein called vitellogenin as a major nutrient source. The authors looked for the genes associated with the production of vitellogenin, of which there are three in the chicken. They found that while monotremes still have one functional vitellogenin gene, in placental and marsupial mammals, all three have become pseudogenes (regions of the DNA that still closely resemble the functional gene, but which contain a few differences that have effectively turned the gene off). The gene-to-pseudogene transitions happened sequentially for the three genes, with the last one losing functionality 30-70 million years ago.

Therefore, mammals already had milk before they stopped laying eggs. Lactation reduced dependency on the egg as a source of nutrition for developing offspring, and the egg was abandoned completely in the marsupial and placental mammals in favor of the placenta. This meant that the genes associated with egg production gradually mutated, becoming pseudogenes, without affecting the fitness of the mammalian lineages.

We knew that milk arose before the split between the mammal lineages. In all probability the gene for casein problably arose at the younger end of their projected chronology. However, there's the possibility that this arose much earlier in the synapsid line.

Just imagine: lactating Dimetrodons. boggle.

It would blur the line between the various therapsid lines if it did hold true. Now if we found evidence of hair too...


Zach said...

I think you're misunderstanding the paper's conclusion. It says that the common ancestor of Mammalia was lactating. Earlier dead-end stem groups may not have. And non-mammalian synapsids? I somehow doubt it!

It certainly would add to the "pigginess" of lystrosaurs, though, wouldn't it?

Will Baird said...

The paper is saying that lactation predates the split of the different mammal lineages. The time frame that they discuss is between 310 to 200 million years ago. Obviously, if its 200 million years ago then its strictly a mammalian trait. However, the paper also gives a chronological bound of 310 million years ago for the genes that code for the protein to have arisen. Obviously, it's unlikely to be 310 mya. However, that's a wide range and there were multiple points which this could have arisen between the bounds, Zach.

I was speculating that it was premammalian since mammals are only extent during a fraction of that time period bounded.

The Dimetrodon comment was meant as a (o.0) mind warping visual. The remainder of my comments were to point out that we'd need to rethink a few things if hair and lactation predate our mammalian branch point. It would make things...fuzzier.

Sorry if I was unclear.

Zach said...

Kinda throws a wrench into our birthing strategies for the walrodont, doesn't it?

Will Baird said...

Depends on when live births happened. The assumption is that the monotremes are retaining the primitive condition (egg laying).

AFAIK, there have been no synapsid eggs ever found. Or monotreme for mammals. It IS obvious that milk giving was pre abandonment of the egg (assuming egg laying as basal). probably won't screw us too badly.

There is just so much we don't really know.

Anonymous said...


Will Baird said...

ok, Carlos, what was that. You, uh, never do something of that nature.

You're making me nervous, man.

Anonymous said...

It's a sweet result. People have tried dating mammalian divergence using casein before -- if memory serves, the first research along those lines used paper electrophoresis, the bearskin and stone knife level of macromolecular analysis. However, casein is not a good candidate for a molecular clock for a variety of reasons.

But I would never have thought about looking for a vitellogenin pseudogene in mammals, although in retrospect it's obvious.

Will Baird said...

So how would you go about narrowing the error bars there? I mean 110 million years is a doozie of set of error bars.

I would guess that it would be through trying other genes or pseudogenes to build up an overlapping set of chronospaces.

Then correlate with fossils?

Just guessing here.

Anonymous said...

The dating is tree independent, which strikes me as a little unusual. They used the functioning vitellogenin sequences from the chicken, and ran it through batches of a neutral evolution simulation (that is, no selection) to get the dates -- 100K runs at 10 Mya intervals. It's Figure S4 at the PLOS website.

The large error bars appear to be inherent to their method. If I am understanding it correctly, the range of years comes from fitting the number of stop codons found in the sequences to the simulated distribution. But because of the timespan, the sequences are already not that far from the entropic limit. Hence the wide range of estimates.

It's like, um, knowing an original piece of text has a only a single letter 'x'. It's then subject to a random but neutral process of changing any letter. In the long run, the final frequency count for the text is going to be 1/26 for x, about 38 out of a thousand.

If the original text had one x out of a thousand, and the divergent one had 10, you're going to get a much better estimate of the divergence time than if the divergent one had 34 x's out of a thousand. From what I can tell, the mammalian vitellogenin pseudogenes are closer to the 34 end.

The way around it: more sequences and stronger statistical analysis.