Anyway, I was reading Sean Hogan's book Trees for All Seasons, and there I encountered a gardeners' myth that refuses to die: that our favorite unusual rose-family trees are closely related to each other. It's an attractive thought, that Polylepis and Leucosidea and Lyonothamnus and Vauquelinia are part of the same group of garden-worthy trees that somehow got scattered across the globe. Unfortunately, it's also wrong.
I thought I'd blog about this topic, as it not only fits nicely with my research interests (essentially the interface of geography, non-floral form, and evolution, should you want to know), but also provides good system in which to discuss evolution and why we change names. Be prepared for vocabulary overload.
The first task is to define what we mean by a group. We tend to think of things as sorted according to their form (for which we use the much more impressive-sounding word "morphology"), but we can sort things by several characters (a character being a single feature of morphology). For example, we could group things with red berries together, but then we would have to group hollies and tomatoes together, to the exclusion of potatoes. Clearly, though, tomatoes are more closely related to potatoes than to hollies. So some characters are less important than others, and/or are superseded by the sum of all characters.
And note that we already think of things as related or not. It could be argued that this is simply predicated on membership within our group, but really, we're already inferring that meaningful morphological similarity reveals a common evolutionary lineage. (This concept, that a given character in two organisms is the result of a single change that occurred before they separated, and was passed down to them, is called homology. The opposite, where similarities arose independently, like the fruit redness in hollies and tomatoes, is called homoplasy.)
The same principle of homology is used for molecular characters (we use loads of them, but especially DNA sequences) for computing our hypothesized evolutionary histories. When using DNA, we often try to use regions that don't get used to make proteins, which control the morphology. Because they don't do anything, they presumably don't have any selection pressure on them (in other words, "survival of the fittest" doesn't apply, because they don't affect fitness; this assumption may not be justified across the board). Because of this, evolutionary change in those regions of DNA is more random (this cuts down on large-scale homoplasy, and allows for molecular dating).
But how similar do things have to be to group them under one name? And what if, in our reconstructed evolutionary history, there is a whole branch that is united by nice, easy characters to examine, except for one little twig in the middle that doesn't share our chosen characters? We need an unambiguous criterion.
These next terms will be easier to explain with a diagram.
 |
| Branches and boxes: red, tree; yellow, shrub; orange, can't tell between tree or shrub; green, subshrub (perennial with woody base); blue, herbs that aren't woody at all; purple, unknown. Arrows and dashes: red arrow, trees primarily discussed in this post; light blue arrow, fern leaved shrubs discussed secondarily; blue dot, plants discussed . . . tertiarily; pink dash, genera native in at least in part in western North America, including Mexico; pink dots, genera native elsewhere in North America. Note that reconstructions of plant habit along branches was not computed, and was assigned by me, with no statistical calculations. Unequivocal colors were drawn up branches only when all descendents were the same, or when overwhelmingly similar (again, no stats). Also note that the two species of Prunus on the tree are used to represent the two subgenera of the genus found in North America; neither P. laurocerasus nor P. domestica are native to this continent. |
It's not essential vocabulary, but notice that the bottom two names are from a closely related family, the Rhamnaceae. These are the outgroups (or, since they're on a single branch, a single outgroup). They allow us to determine where the base of the rose family is. Because we know where the base of the tree is, we can say that as we get farther away from it, and toward the branches (which direction I'll call down, as opposed to moving up towards the base... confusing, I know), the branching points (nodes, just like on a plant) get younger in age. Because of the way the tree is drawn, however, we can't compare the age of nodes on different branches; their relative position is changed for ease of reading. The topology, or order of branches, remains the same.
Think of it as an overgrown family tree (in the other sense...), and it'll be much easier to read.
(I'll ignore that there are different types of these diagrams that mean slightly different things. For this post, it's just a "tree," a visual representation of an evolutionary hypothesis. Most trees also have statistics on the nodes that show how well the data resolve into a given topology.)
Any group of branch tips can be described by one of three terms. A group that includes a common ancestor and all of its descendents is described as a monophyletic group (we also call them clades, because it's a shorter word). An example of clade is near the bottom: the group of Cercocarpus, Chamaebatia, Cowania, Purshia, Dryas, and up the tree at least to the point where all those branches come together. If the group contains a common ancestor but only some of its descendents (for example, the same group, but we just don't like Chamaebatia, because of its ferny leaves), we call it paraphyletic. If it includes descendents but no common ancestor (think the first group, but we decide to add Chamaebatiaria and Spiraeanthus, because they have ferny leaves like Chamaebatia), it's polyphyletic. The latter two can run together if people aren't very careful with their words.
(One last term: the two group that branch off a given node are called sister to each other. In this case, the outgroup is sister to the rest of the tree.)
Now consider again the question above: how similar do things have to be to put them into a named group, and what if there's a weird thing wedged in the middle? Try as some might to assign a statistical cutoff of similarity to group things together, or whether or not things can interbreed, there's really only one way to distinguish groups that's applicable across the board: only monophyletic groups get names. There are cases wherein this criterion stretches (such as when hybrids stabilize and start hybridizing themselves), but it has yet to break, and there's no foreseeable situation wherein it would. It's another problem, which goes back to the basic question itself, that some genera may encompass more variation than others, but with our criterion, the question is simplified from "how do we justify grouping these together" to "which node do we want to use." So it's a simple judgment call, and no longer a theoretical question.
Consider that Cowania is often now lumped into Purshia - basically, the former just has larger flowers. We can do that with few repercussions because both genera are small, so we have few new names to publish, and because the two groups are sister, so there's nothing wedged in between.
As another a common example, consider the genera Photinia and Heteromeles, which I've indicated by light blue dots to the right of the name. The former includes the overly-common P. x fraseri, as well as some worthier plants. The latter, tragically underused, was formerly included in Photinia, and some gardening references still claim that combining them is a valid option, if you overlook their differences and focus on their similarities. But if we want them in the same genus (and assuming the hypothesis above correctly reconstructs evolutionary history), we need to include their common ancestor, and all of its descendents. Then, because there are also rules for names, we'd need to use the oldest valid or conserved genus name within the group, or conserve a name we prefer (that is, propose to the International Botanical Congress that the rules be overriden in favor of using a younger name, generally because the older name was published in an obscure place and never caught on).
So, if we want Photinia and Heteromeles to be in the same genus, the oldest genus name among the descendents of their common ancestor is Pyrus, and the new concept of the genus would include not only pears, but also apples, cotoneasters, loquats, Stranvaesia (also, and more pervasively, sometimes included in Photinia...), and the true quince. So please, please, please stop saying that including Heteromeles (or Stranvaesia) in Photinia is valid. Because if that's what you want, all that other stuff has to be in the genus as well, and you'd better get on that proposal to conserve the name over Pyrus, Malus, and the others; the next Congress is in 2017.
* * *
Now on to the trees. The ones commonly claimed to be closely related are marked on the tree with red arrows. Note how far apart they are... This will be much easier with some names.
Rosaceae is generally divided into three subgenera. First is Rosoideae, which is the first major branch, and encompasses about the top third of the tree. The next is Dryadoideae, which is the group with Dryas I used as an example above. It's sister to the Amygdaloideae, which forms the rest of the tree. The three are distinguished fairly well by ovary morphology, but as you might expect, there's quite a bit of variation within Rosoideae and Amygdaloideae. The topology between these three subfamilies presented here may not be correct, as may some of the major clades within them. More recent studies indicate that Dryadoideae and Rosoideae may be sister. The genera included in the subfamilies, however, remains the same.
(Fun fact: the cherries and plums and other fruits used to be a separate family, Amygdalaceae, but much of what's in Amygdaloideae, like Kerria and Spiraea, as well as the Dryadoideae, were already included in Rosaceae. Even without DNA evidence, it was fairly obvious that the Rosaceae was then (quiz word). So Amygdalaceae was subsumed into the Rosaceae. Robert Frost, always an angry and unpleasant little man in real life, for once let his frustration spill out into a poem.)
Notice that Lyonothamnus and Vauquelinia are in Amygdaloideae, while Polylepis and Leucosidea are in Rosoideae. And Polylepis and Leucosidea, while both in the tribe Sanguisorbeae (the clade at the top from Acaena to Spenceria), are in different subtribes (Sanguisorbiinae and Agrimoniinae, respectively). Also, while both are trees that have pinnate, hairy leaves, consider that most of Rosoideae has pinnate, hairy leaves, and some woodiness appears throughout that subfamily. (I won't delve into wood anatomy here. Keep repeating that.)
Leucosidea is, however, sister to Hagenia abyssinica, an interesting tree that would probably do well in northern California.
Lyonothamnus, even in the most recent study (published March of this year), is found to be sister to the rest of the Amygdaloideae. (Actually, I wouldn't mind having it in its own monotypic subfamily, but it's fairly similar in fruit to some other Amygdaloideae.) In his book, Hogan discusses Lyonothamnus and Vauquelinia as if they were two variations of the same thing, in so many words, as if they were sister to each other. While Vauquelinia is another isolated lineage, it's nested well up in the Amygdaloideae. (Did I not introduce nesting? That just means that it's included within a specified larger clade.) In fact, the split between Lyonothamnus and the rest of the Amygdaloideae has been dated to the Cretaceous. So they're not two sides of the same coin.
Furthermore, Lyonothamnus isn't an example of island gigantism, as Hogan claims. Besides the fact that his conclusion is drawn from the presumed relationship to the shrubby Vauquelinia, the fossil record doesn't support this, either. Granted, of the four fossil species of Lyonothamnus, the largest-leaved had leaves only about half the size of the of the modern species, including those leaves reconstructed from fragments. Leaf size, however, can evolve fairly quickly. More telling are the fossilized trunks, and the past ecology of the area (most fossils are from western Nevada). The local flora from the same period of time includes hemlocks, maples, and Zelkova, and the fauna included both beavers and camels. And rhinoceros.
(Another fun fact: the earliest fossils assignable to Lyonothamnus are about 23 million years old, and were found in the Cascades of Clackamas County, Oregon.)
(Last fun fact: there were rhinoceros in Washington until about 2 million years ago.)
That may have sounded a bit rantier than I intended, but hopefully the science behind all of this is a bit clearer. Plus, the science is just awesome.
(Even botanists make the same kind of errors. The genera indicated by blue arrows all have ferny leaves, which led some workers who found ferny rose-family fossils to decide that Chamaebatia was sister to Chamaebatiaria. The different types of fossils, however, if you consider the veins that point to the sinuses in between the teeth, may support there being two different ferny lineages from a long way back. It's ongoing work, though, so don't any definitive repudiation just yet.)
No comments:
Post a Comment