Deep Homology

Darwin referred to the products of evolution as ‘endless forms most beautiful’. Disparity (the word meaning “diversity of morphology”) can be classified: “All mammals have titties” is a valid biological statement. But within all this disparity lie unifying characters: all mammals have boobs, but they also have a backbone; frogs have a backbone too, so that’s a common trait. But a tadpole (baby frog) does not look like a frog at all; developmental biology and ontogeny also greatly increase disparity.

Even within groups, you may get incredible variations and twists: think of how different insects and barnacles are from the rest of the crustaceans; the mites from the rest of the chelicerates; the cephalopods from the other molluscs; the echinoderms from other bilaterians; the cetaceans from other mammals. These are all examples of how free evolution is to modify the underlying ground plan. And that’s the topic of this post: how does all this variation arise?

When someone colloquially talks about wings, they equate the wings of an insect and of a bird, uniting them because they perform a similar function. For evolution, this is obviously false: under that concept, kiwis and emus do not have wings. It’s an extreme example, but it demonstrates my point perfectly: when speaking of morphology, function plays no role. The reduced forelimbs of the flightless birds are homologous to those of ducks, chickens, pigeons, etc. But they have nothing to do with the wings of a fly. We know this from developmental biology. That’s where our discussion will play out: in the embryo, its cells and its genetics.

The example above could be extended to the compound eyes of arthropods vs. lens eyes of vertebrates, or the ears of Drosophila and the ears of a human, or the ‘head’ or the ‘brain’: same function, different origins. But when looking at the developmental genetics underlying their formation, we notice that they use the same/similar pathways to generate the structures. For the eyes, two genes (Pax-6 in vertebrates and eyeless in insects) are expressed in those animal taxa and play an integral role in the formation of photosensitive cells (not eyes – remember the false comparison!), whether it’s in an arthropod or in a mouse. This concept, that there are similar underyling genetic mechanisms leading to similarly-functioning structures among non-related clades is called ‘deep homology’.

Note that, much like ‘species’, ‘homology’ does not have a set definition and varies depending on your own disciplinary bias. I use it in the broadest, colloquial meaning: a structure/genetic sequence/developmental pathway/etc. derived from a last common ancestral source. In other words, a homologous structure is a synapomorphy (my disciplinary bias is phylogenetic).

Let’s look at a classic example of deep homology. Polychaete parapods, velvet worm lobopods, ascidian ampullae and echinoderm tubefeet (all appendages/outgrowths from the body wall, used as legs) have the same gene expressed at their tips during development: Distal-less. There are two explanations for how this can come about. It could be convergent, in that the gene got co-opted to have a function in appendage development at least four times independently during evolution. Or that this function was present in the last common ancestor of all those groups – in this case, the last common ancestor of the bilaterians (the ‘Urbilaterian’). The latter has numerous implications, the first and foremost being that the Urbilaterian must have had similar appendages.

Whether this is true or not we’ll see at the conclusion. Let’s look at another example: segments. The most speciose and disparate animals have a segmental body plan: look at the arthropods, chordates and the annelids; it obviously has an extremely high adaptive value. This type of body is where several units (where unit can mean a tissue, organ, cell type, etc.) is repeated serially from the front to the back of the animal. In developmental biology, the ‘front to back’ direction is called the anteroposterior (AP) axis, and that’s how I will refer to it from now on. Basically, the animal is made up of stacked together modules, and each module can then be freely modified by evolution to suit the particular ecological needs of the organism. Compare a spider with an insect: they both come from the same segmented proto-arthropod, but while the ground plan is the same, evolution has produced a bewildering amount of variation.

Let’s dive into segmentation a bit deeper. What differentiates a segmented animal from a non-segmented one? First, the anterior and posterior of the animal (the front and back) get defined by specific proteins that affect differentiation and cause slight asymmetries in the egg. Then, once the two poles are established, the whole axis is extended; this occurs from the posterior end. Up until now, we have no differences between segmented and non-segmented animals: this occurs in all bilaterians, often with the very same highly-conserved molecular genetic mechanisms. The final step is the decisive one: this is when the segmentation pattern is pushed onto the elongating AP axis, causing the formation of segments.

In vertebrate embryos, undifferentiated mesodermal tissue (called the pre-somitic mesoderm, PSM) gives rise sequentially to somites – by sequentially, I mean along the AP axis by terminal addition (i.e. new segments are glued on at the back). Somites are the individual segment units of vertebrates. The process is controlled by a cascade of gene expression, in turn orchestrated by the Notch signalling pathway. This is extremely predictable and works like clockwork. Arthropods have a somewhat similar system; they don’t have a PSM, but they do have a segmentation zone. They also use Notch signaling, and is also clocklike in its precision – the similarities with the vertebrate system are quite remarkable. Segmentation hasn’t been studied enough in annelids, but it also proceeds along the AP axis; there is no generalised knowledge of the molecular genetic systems underlying it though, so we will not consider them further.

There are other conserved ‘segmentation genes’ between the two phyla (chordates and arthropods). The engrailed gene is one example; developmental biologists working with arthropods detect it to mark the limits between embryonic segments. In chordates, engrailed is also found. In vertebrates though, it mostly has roles in patterning the nervous system. But in amphioxus, the animal regarded as being the most basal chordate, engrailed is expressed in much the same places as in arthropods. We can thus speculate that engrailed was ancestrally involved in segmentation, but got co-opted in the vertebrates for another function.

I hesitate to discuss the ancestral state of segmentation because it is central to the question of bilaterian phylogeny. And as you all know, I am very opinionated and in this case, I don’t want to “officially” take a stance (even though I do have one personally), so I will keep this simple without going into caveats. The last two paragraphs summarised the main points speaking for a common origin of segmentation: if arthropods and chordates use the same developmental genes to make their segments, then extrapolating back to their last common ancestor, the proto-Bilaterian, shows that it was segmented too. But that is a very narrow minded view: the chordates and arthropods may be the 2 largest phyla, but they are just 2 branches out of ~30. Saying segmentation (a very successful adaptation by any means) was lost 30 times is simply not parsimonious (parsimony = Occam’s Razor in phylogenetics). On the other hand, saying it emerged twice independently requires many less changes and is preferred; but note that loss of segmentation has been observed in many successful groups, such as in sipunculids and echiurans (various annelids). The change is associated with change in lifestyle (e.g. parasitism or sessile life).

There is also the fact that in chordates, segmentation arises from the mesoderm, whereas in arthropods and annelids, it comes from the ectoderm. Chordates have no clear boundaries between their segments, whereas arthropod segments are clearly separated by the cuticle and annelid segments by septa. Also, annelids derive their segments from special stem cells (teloblasts), a unique method not found in the arthropods or the chordates. This all speaks against a common proto-Bilaterian origin of segmentation.

As I said, I will not state my own opinion on this, I’m just telling all the facts here.

Just for fun, let’s look at the development of the brain. The head is the command center where you have the most sensory inputs (eyes, ears, antennas, etc). The brain is always found in the head because of convenience – their development is not really bound together. When discussing segmentation, we focused on the AP axis. Here, we’re exclusively at the anteriormost region of the animal, so if anything, the dorsoventral axis (DV) will be mentioned.

The nervous system arises from the ectoderm, as does the head. Both arthropods and vertebrates have a single gene (decapentaplegic (dpp) in arthropods; Bone morphogenic protein 4 (BMP4) in vertebrates) taking care of non-neural ectoderm, and a single factor for differentiating the neural parts (short gastrulation (sog) in insects; chordin in vertebrates). The unique thing about chordates though is that their DV axis is inverted – that’s why the human spine in ‘on your back’ (now you see why AP and DV are much more practical to use).

Once the neural ectoderm has been specified, so-called columnar genes divide it into three parts, each part specifying a different type of neural precursor cell (you can’t call them ‘neurons’ yet, as there is no nervous system at this stage of the embryo). And again, these columnar genes are highly-conserved between the arthropods and the vertebrates, in much the same situation as the segmentation genes we looked at above. And even within the three ‘columns’, the genes and factors that work in there are also highly conserved.

However, there is a significant difference between arthropods and vertebrates: forming the neural tube (precursor to the spine) in vertebrates and the formation of the ventral nerve cord of arthropods. In vertebrates, you have a sheet of ectodermal-derived neural cells called the neural plate, which then undergoes neurulation. This is basically when it folds in on itself because of gastrulation. At the end, you get a neural tube with epidermal ectoderm on top (skin) and the notochord at the bottom. In insects, nothing of this sort happens. There, you have a layer of undifferentiated ectodermal cells. A specific signaling pathway singles out neuroblasts (neural stem cells), which then split off from the rest of the layer and move further inside the embryo, where they divide asymmetrically, making the ganglion mother cells (GMC), which divide further to produce neurons and glia.

Now these couple of examples are a bit misleading. In each organism, there are four levels in a hierarchy: genes, development, morphology and behaviour. But there is a hierarchical disconnect: non-homology in the lower levels does not necessarily mean that the higher levels will also not be homologous. The best example for this is salamanders and the rest of the tetrapods. In all tetrapods (except salamanders), the embryonic digits (fingers, toes) have tissue between them, which later on dies (by controlled cell death). In salamanders, this is different: each digit develops individually. But obviously, the digits have the same evolutionary origin (from looking at any phylogenetic scheme); there is a hierarchical disconnect between the developmental pathway and the morphology: the development is not homologous, but the end morphology is. The same can happen with new genes causing homologous morphologies or different morphologies causing homologous behaviour. This is why there is no set definition for ‘homology’.

And this is also why those examples are misleading. As I said, vertebrate and arthropod eyes are not homologous, despite sharing the same developmental pathway. Pax-6 and eyeless are othologues – they are the same tool – but they are used in very different contexts (the embryo of a chordate is very different from that of an insect). Looking more basally, Pax relatives are also found in cubozoans (medusae with a complex camera eye) – but there, it plays no role in the eye’s formation. This hints at convergence. Going back even further, sponges and placozoans also have Pax genes, as well as other genes involved in eye development in arthropods and vertebrates. The point is that while those master regulatory genes are used to guide eye formation, they (and their interactions!) existed long before the appearance of anything resembling eyes.

That said, bringing up that matter of definition is rather pedantic: the point of this post still stands. There are core genetic pathways that are conserved through all of life – the more conserved, the more essential. The hindwings of butterflies and mosquitoes are homologous, despite having very different functions: in butterflies, they are wings, in mosquitoes they are a sensory club. There is one gene that controls the formation of an outgrowth at the segment of the body (Ubx), but the later modifications are controlled by other genes and other factors. But they are homologous by the very fact that they come out of the same place on the insect body, and the same master gene causes their formation.

But what about something like sleep? Birds and mammals sleep, and both undergo slow wave sleep and rapid eye movement. Both clades show similar physiological regulatory mechanisms. Sleep is conclusively convergent in them, as is their intelligence. However, they also share common cellular and genetic pathways to achieving sleep. And this goes even deeper: flies and bees do it too, and also share the same mechanisms, as do the cubozoans. But here’s the catch: all these animals are sophisticated animals with big brains and a propensity for complex and intelligent behaviour. Sleep is but a side-effect of this – that they employ the same processes to achieve it is nothing more than coincidence (i.e. convergence), especially since these processes and the genes enabling them have existed long before these animals evolved. Simply sharing the same pathways does not point to homology, it only points to a gene that is suitable for that function and is thus favoured for convergence: in a way, it is a case of “predictable evolution”.

So deep homology, what we’ve discussed so far, involves co-option of already existing genetic architectures and/or changes in regulatory mechanisms, and this is arguably the largest origin of novel structures. But let’s not forget the role that completely new genes play! Duplications of sequences within the genome are common, producing one extra sequence free to be given functions (or not, it may just become redundant). Even without duplications, new genes can arise from previously noncoding sequences by mutation. But that’s a post for another time.


  1. Pingback: Going Down the Rabbit Hole of Evo-devo: Deep Homology • theGIST

  2. Pingback: Science communication in the biomedical sciences: challenges, opportunities and new approaches | Gedankenexperimente

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.