Making trees: how are mammals related?

Figure 1.  An 1837 sketch from Charles Darwin's notebooks.

How do scientists tell what organisms are related to other organisms?  It's not easy to go back in time and visit with the common ancestor of chimpanzees and humans for example, so we need some approaches to place organisms into natural groups.  This was precisely what the Swedish biologist Carl Linnaeus first attempted in a 12 page publication called  Systema Naturae in 1735.  By 1758, the tenth edition included descriptions of 12,100 plants and animals.

Modern biologists are just as concerned with appropriately grouping related organisms as was Linnaeus.  Since the time of Darwin we have used tree diagrams to represented the relationships among related organisms (Figure 1).  Humans and chimpanzees for example can be placed on the tips of branches that can be followed back to a larger branch that represents the ancestor we share in common.  To see how we are related to monkeys, bats, dogs, or sharks, we'd have to track back closer and closer to tree trunk and then trace paths back out along more distant branches.  The length of that branch-tracking journey represents the distance of the relationship between any two organisms on the tree.

When I started teaching evolution at Saint Michael's College I wanted an authentic activity that students could complete to make an actual evolutionary tree.  It's quite frankly boring to simply study trees completed by others and memorizing the branching patterns seems utterly pointless.  My quest was for a prepared procedure that would walk students through the actual process used by evolutionary biologists to make the trees that we call phylogenies.

Figure 2.  Data from 2015 Saint Michael's College students.

My quest was fruitless and so I wrote my own student-accessible procedure based on the techniques used by professional biologists.  My students use a collection of real and replica mammalian skulls.  They come up with a laundry list of observable skull characteristics: canine teeth, forward facing eyes, snouts, notches, grooves, and whatever traits they can come up with and clearly communicate to their class mates.  Each skull is then scored for each trait: "1" represents "present"; "0" represents "absent".  A subset of this year's data is shown in Figure 2 and from it you can see that bobcats, and lynx have tiny molars and Y-shaped premolars that are not found in the other organisms on this subset of the larger list.  These unique traits together with many others can be used to infer that bobcats and lynx are related.

Figure 3.  The arbitrary starting tree; click to enlarge.

This year we measured 46 traits from 26 mammals and then placed the mammals on an arbitrarily drawn tree (Figure 3).  I describe the starting tree as "arbitrary" because no attempt was made to place related organisms on adjacent branches: bobcats share a branch with polar bears and sheep for example.  Obviously this is not ideal and more importantly, it suggests that tiny molars evolved bobcats, and again in lynx, and evolved independently three more times in lions, cougars, and Bengal tigers.  It makes more logical sense that tiny molars evolved in some ancient cat and that the trait was passed down to modern cats.  If this is true we should expect to see this traits in other cats and indeed we do see exactly that in the two other species we have in our collection: house cat and caracal cat.  Scientists call this "the principle of parsimony": why invoke 5 evolutionary events when just 1 is needed to explain the data.

We refer to the gain or loss of traits as "transitions" and we used software to count the 251 transitions needed to explain the arbitrary tree.  My students then move branches around on the tree and the software automatically recalculates the number of transitions.The goal is to generate the most parsimonious tree, or the tree with the smallest number of transitions.  This becomes competitive as student groups report out on their shorter and shorter trees during the lab session.  As a homework assignment, the students compare their trees to trees published by evolutionary biologists.  This places their work in the larger context and I have found that the comparisons generally fare very well.  As a result of writing this blog I think I'll ask my students to use their trees, together with published trees to write several hypotheses about the traits of skulls they have not yet seen.  Sounds like a whole new lab!

Figure 4.  A tree made based upon the observed skull traits.

What I don't tell my students is that the software has a feature that does the work automatically.  The software is unbiased; it won't place the polar bears near the brown bears just because that might make most sense.  Instead, the software makes groups that minimize the number of transitions and generates a tree based solely on the data (Figure 4).  This data-based tree requires 121 transitions and places organisms together in ways that very closely match the tree of life, a phylogeny generated by professional biologists.

Importantly, 15 students dreamed up 46 traits without consulting published work and without reference to what some other biologist might think of as a 'good trait'.  They worked in groups and the only criteria for choosing traits were: that they could communicate the trait to their peers; the trait should occur in at least 2 skulls; the trait could not occur in every skull.

This is our third year running this experiment.  We use different skulls each year to keep it interesting.  The list of traits that students come up with changes each year also.  If other teachers would like to run this exercise on skulls or on other organisms you can find all of the needed information in this short paper.

Figure 1 from Wikimedia Commons.  Other figures generated at Saint Michael's College.