Algae have the annoying tendency to show high levels of cryptic diversity, i.e. with distinct species being morphologically indistinguishable. This has been shown repeatedly by first assessing species boundaries using DNA work or crossing studies, and subsequently comparing these species boundaries with morphological features.
I’ve always been interested in how morphological complexity of organisms relates to their tendency to produce cryptic species. When we found cryptic diversity in Pseudochlorodesmis, a genus in which the algal body is utterly simple, we argued that this may be due to its simplicity: “From a strictly morphological point of view, it is simple to conceive that the potential prevalence of cryptic diversity within any given taxon is a function of its morphological complexity. For example, if the morphology of the members of the taxon can be scored as a set of X binary characters, and morphological species boundaries are defined by a minimum of one character difference, the maximum number of morphologically determinable species increases exponentially with the number of characters available (N = 2X). In other words, for a higher taxon containing a given number of species, chances of encountering cryptic diversity increase dramatically with decreasing morphological complexity.” (Verbruggen et al. 2009 J. Phyc. 45: 726-731)
Of course, the N = 2X is a theoretical maximum, and I wouldn’t expect all theoretically possible morphologies to be produced in the course of the evolution of a lineage. To look at this in some more detail, I’ve done a few simulations. This approach consists of generating phylogenetic trees containing a number of species, and subsequently letting a set of traits (i.e. morphological characters) evolve along this phylogeny at a rate that corresponds to those measured for a real algal morphometric dataset. The result of this exercise is a set of values for each trait for each species in the phylogeny. Those can then be compared with each other to evaluate how many unique morphologies there are and how many of the species can be reliably distinguished from one another morphologically.
First, I wanted to quantify how fast your average discrete morphological trait evolves in algae. So I took one of my morphological datasets for Halimeda (mostly unpublished, but similar in nature to Verbruggen et al. 2005 J. Phyc. 41: 606-621) and a corresponding phylogenetic tree of the species in that dataset. The tree is a chronogram, which was rescaled to have a root-to-tip path length of 1. Five of the variables in the dataset are discrete, and I calculated the rate of the Markov process for these using the fitDiscrete function in the geiger package for R. Here are the results:
> print(mkr) perwall perfusions secinfl secper segundul 5.0208161 1.0331563 0.8157761 5.4067015 0.1252521
Cool. The evolutionary rates of the traits vary quite a bit. I decided to start the simulations with the lowest of these rates (I used 0.1), and then increase the rate later on.
Here’s a breakdown of the simulation function:
- simulate Yule tree (pbtree from phytools package)
- rescale tree to have root-to-tip length of 1
- simulate the evolution of the desired number of traits along the tree (rTraitDisc from ape package)
- count the number of unique morphologies (trait combinations) produced during the simulation
- loop through all species pairs and score how many are distinguishable from each other (two species are considered distinguishable if they have at least one trait that differs between them)
This procedure was repeated for trees containing different numbers of species (from 10 to 400), with the number of unique morphologies and the fraction of distinguishable species pairs being retained at each step. Now let’s plot some results…
That’s quite spectacular. At this rate of trait evolution you get MUCH fewer unique morphologies than there are species. For organisms with 20 traits, you get only about 50 unique morphologies even though there are 400 species. That’s a lot of cryptic diversity. For organisms with 10 traits, the situation is even worse and only about 20 unique morphologies are produced for 400 species. Okay, now let’s plot the percentage of distinguishable species pairs…
The blue triangles (20 characters) clearly lie above the golden dots (10 characters), reflecting that organisms with lower morphological complexity have fewer unique morphologies and are harder to distinguish from one another. In other words, lower morphological complexity leads to higher levels of cryptic diversity. That was expected, but nice to see it confirmed in the simulation.
Another interesting feature of this graph is that there is no relationship between the number of taxa in the tree and the percentage of distinguishable species (flat lines). At first, this seemed counterintuitive to me. When given a certain amount of time to diversify (one time unit from root to tips), and with a fixed rate of morphological evolution, shouldn’t more diverse lineages have more species that look the same? Actually, no, because trees with more species have a higher total tree length. The root-to-tip distance is still the same, but you have more lineages that add to the total tree length and thus to the total amount of evolution in the morphological trait. So those flat lines do make sense.
I’ve also tried these simulations with different rates of evolution:
Clearly, traits with higher rates are better at distinguishing between species than slower traits, reducing the number of cryptic species in a lineage.
In conclusion, let me just wrap up the core results from this exercise:
- There are substantially fewer unique morphologies than there are species.
- Character-poor lineages produce fewer unique morphologies than character-rich lineages.
- Lineages with fast-evolving traits feature less cryptic diversity than those with slow-evolving traits.
Because of these results, we can expect cryptic diversity to abound, especially in character-poor lineages. As such, for any given algal taxon, we should expect to be unable to distinguish between at least some and possibly many of its species based on morphology alone.
You can download the code for these simulations from here.