A good portion of what I’ve been teaching so far uses Drosophila as a model system — it’s the baseline for modern molecular genetics. Unfortunately, it’s also a really weird animal: highly derived, specialized for rapid, robust development, and as we’ve learned more about it, it seems it has been layering on more and more levels of control of patterning. The ancestral system of establishing the body plan was far simpler, and evolution has worked in its clumsy, chance-driven way to pile up and repurpose molecular patterning mechanisms to reinforce the reliability of development. So I promised the students that this would be the last day I talk about insects for a while…we’ll switch to vertebrates so they can get a better picture of a simpler, primitive system. What we’ll see is many familiar genes from flies, used in some different (but related!) ways in vertebrates.

But today I continued the theme of epistatic interactions from last week. Previously, we’d talked about gap genes — genes that were expressed in a handful of broad stripes in the early embryo, and which were regulated in part by the even broader gradient of bicoid expression. The next level of the hierarchy are the pair rule genes, which are expressed in alternating stripes — 7 pairs of stripes for 14 segments.

First point: notice that we are seeing a hierarchy, a descending pattern of regulatory control, and that the outcome of the hierarchy is increasing complexity. One gene, bicoid sets up a gradient that allows cells to sense position by reading the concentration of the gene; the next step leverages that gradient to create multiple broad domains; and the pair rule genes read concentrations of gap genes and uses the boundaries between them to set up even more, smaller and more precise domains of stripes that establish the animal’s segments.

This is epigenesis made obvious. The 14 stripes of the pair rule genes are not present in the oocyte; they emerge via patterns of interactions between cells and genes. The information present in the embryo, as measured by the precise and reproducible arrays of cells expressing specific genes, increases over time.

So part of the story is hierarchy, where a complex pattern at one stage is dependent on its antecedents. But another part of the story is peer interaction. Cells are inheriting potentials that are established by a cascading sequence of regulatory events, but in addition, genes at the same approximate level of the hierarchy are repressing and activating each other. We can tease those interactions apart by fairly straightforward experiments in which we knock out individual pair rule genes and ask what the effect of the loss has on other pair rule genes. I led the students through a series of epistatic experiments which started out fairly easy. Knock out a pair rule gene that is expressed in odd numbered parasegments, for instance, and it’s complement, the pair rule gene expressed in even parasegments, expands its expression pattern to fill all segments. Sometimes.

Some of the experiments reveal simple relationships: hairy suppresses runt, and runt suppresses hairy. That makes sense. They have mutually exclusive domains, so it’s no surprise that they exclude each other. But then we looked at other pair rule genes which are expressed in patterns slightly out of phase from the hairy/runt pair, and there the relationships start getting complex. Genes like fushi tarazu are downstream from all the others, and their effects are straightforward (their loss doesn’t disrupt the other pair rule genes), but genes like even-skipped have much messier relationships, and the class was stumped to explain the results we get with that deletion.

So I asked them to come up with other experiments to tease apart these interactions. I was somewhat amused: when I think along those lines, I come up with more genetic crosses and analyses of expression patterns — I think about regulatory logic and inferring rules from modifications of the pattern. Students nowadays…they’re so much more direct. They want to go straight to the molecular biology, taking apart the genes, identifying control elements, building reporter constructs to see gene-by-gene effects. I felt so old-fashioned. But we also had to talk about the difficulty of those kinds of experiments, and that often the genetic approach is better for building a general hypothesis that can be fruitfully tested with the molecular approach.

Then we stopped — we’ll come back to flies later, and start looking at some specific subsets of developmental programs. Next, though, we’re going to take a big step backward and look at early events in vertebrates and progress through that phylum until we see how they build segments. I’m hoping the students will see the similarities and differences.

Slides for this talk (pdf)

Today we talked about gap genes and a little bit about pair rule genes in flies, and to introduce the topic I summarized genetic epistasis. Epistasis is a fancy word for the interactions between genes, and we’ve already discussed it on the simplest level. You can imagine that a gene A, when expressed, activates the expression of gene B. The arrow in this diagram? That’s epistasis.

So far, so simple. This could describe how bicoid activates zygotic hunchback for instance. But of course not all epistatic interactions are linear and one dimensional; often one transcription factor will turn on or repress multiple genes — so A might switch on genes B, C, and D.

But wait! Now there is the potential for all kinds of combinatorial interactions: maybe C has positive feedback back on A, and B activates D and C, and D activates B, and C represses B. There’s a whole mathematically bewildering world of possibility here.

And it gets worse and worse. B, C, and D could have downstream effects on other genes, like E, F, G, and H, and each of those interact with each other and can have feedback effects as well. It’s not at all uncommon to be taking apart the sequence of events of a developmental pathway and discover a whole tangled snarl of epistatic interactions that lead to complicated patterns of gene expression.

And that’s molecular geneticists and developmental biologists do: they try to tease apart the snarl, asking how each gene interacts with all the other genes in the system, working out the kind of genetic circuitry shown in those diagrams. Often the approach is take it one gene at a time: knock out F, for instance, and ask what happens to the expression patterns of A, B, C, D, E, G, and H. Or upregulate D, and ask what all those other genes do. If you like logic puzzles, you’ll love epistatic studies, because that’s what they are: grand complicated logic puzzles with multiple cascading effects and usually only partial knowledge about what each component does. You’ll either have great fun with it all, or cultivate great headaches.

So most of the class hour was spent going through examples of these puzzles. The gap genes, for instance, are expressed in broad stripes in the embryo, and we can try to decipher the rules that establish the boundaries by taking out components. If hunchback is deleted, what do the giant, krüppel, and knirps stripes look like? Take out krüppel, what happens to knirps? So I led them through this series of experiments, asking them to come up with general rules regulating the expression of each stripe, and then using those rules to predict what would happen if we did a different experiment. I think they mostly got it.

But of course the discussion today was mostly about the gap genes, which are the second tier of genetic interactions (analogous to my third figure above). Next I introduced the pair rule genes, the third tier, rather like my fourth diagram. These are genes that are expressed in alternating stripes corresponding to parasegments in the fly…so we’ve gone from a few broad stripes to many narrow stripes. Each of those stripes, too, is independently regulated, with distinct control regions for each.

The real nightmare begins in the next class, when we start taking apart the many ways all of the pair rule genes interact with each other, and how their position is established partly by regulation by the gap genes and partly by mutual sorting out with combinations of activating and repressing interactions. It’s going to be loads of fun!

Today’s slides.

My students get a full exposure to the Sean Carroll perspective in his book, Endless Forms Most Beautiful, and I’m generally pro-evo devo throughout my course. I do try to make them aware of the bigger picture, though, so today we had an in-class discussion/’debate’ (nothing so formal as a debate, and it was more a tool to make them think about the arguments than to actually resolve a question). Fortunately, there’s one really easy exercise we can do in developmental biology, because some big names in the field have already clearly laid out their positions in a couple of relatively succinct papers, so I had a shortcut to bring the students up to speed on the issues. I split the class on Monday, having half read a paper by Hoekstra and Coyne on “The locus of evolution: evo devo and the genetics of adaptation” (pdf), which argues for the importance of trans-acting mutations in evolution, and another by Wray on “The evolutionary significance of cis-regulatory mutations” (pdf), which argues for the importance of developmental changes through changes in cis regulatory regions.

I drew this little cartoon on the board to illustrate the situation: that changes in the coding regions of genes produce mutations that can have broader effects throughout the cell (trans: they can affect other genes not on the same chromosome), while changes in regulatory DNA will have discrete effects on just the gene on the same strand of DNA they are (cis).

Then I asked them put together an argument as a group advocating for the significance to evolution of their ‘side’, cis or trans, which they then delivered to their opponent, with opportunities for rebuttal and counter-rebuttal.

Ah, pitting the students against one another…always the fun part of teaching.

There was good friendly discussion. Both sides had to dig into their respective papers to find the arguments, and then restate them to make their point, both of which are good exercises. The battle waged to and fro, and then our hour was up and I asked them to vote for who ‘won’, in the subjective sense of making a good argument and persuasively advancing their position. The results:

Which position do you think makes the best case for the significance of their phenomenon in evolution?

Team trans: 1
Team cis: 0
Both positions are important: 8

Minnesota mildness for the win!

I did think one student comment was perceptive and exposed the whole argument for a sham. If they were to go off to graduate school in developmental biology, they wouldn’t be picking Team trans or Team cis: they’d be pursuing a phenotype or a pattern of interest, and then analyzing how it worked and came to be, and they’d simply accept the evidence, cis or trans or both, however it turned out. Follow the data, always.

Now that’s a healthy attitude.

You know I teach the 8am courses every term, right? Every semester for years I get my oddball classes that weren’t present in the curriculum 13 years ago (when I started here) stuffed into the cracks of the schedule. I’m slowly getting to be a little pushier and am gradually making my way into wakier hours with other classes, but so far, developmental biology is still in the darkness. Fortunately, this talk was so jam-packed with excitement and action that they couldn’t possibly sleep through it! Right?

Just a word about the presentation slides: I’m a firm believer that less is more. My goal is not to display my lecture notes, or lists of bullet point slides that make my points for me, but to show complex and interesting illustrations that I talk about and explain — whoa, I know, how radical. I’ve sat through too many talks that flash 60-80 slides at me in an hour, and it’s too much. Take your time, people! That said, I used 18 slides in a 65 minute lecture today, which I felt was a little excessive — I aspire to someday do a lecture with half that number. But I am weak and need the crutch now.

Also, I returned exams today. People asked if I’d post their answers. No way in hell! These are exams and have the privilege of privacy. I will say that in general the students answered well. The goal of that kind of exam isn’t to confront students with a question that has a specific answer, but with a problem that they should explore, defend, or criticise.

So the subject today was maternal effect genes in Drosophila, specifically the prepatterning information that specifies the anterior-posterior and dorsal-ventral axes. Yes! I can tell you’re all excited!

So I gave them the precursor observations to the actual molecular biology, all this lovely modeling of gradients and information domains that was rich with Turing elegance, and then I dashed their optimism with the cold water of reality: molecular biology has shown that instead of beautifully designed systems, we’ve got bits and pieces cobbled together in a functional kludge. Any pretty patterns we do see are the product of brute force coding.

So they got the overall picture of A/P patterning in flies: a gradient of the Bicoid protein, high in front and low in back, is read by cells to determine their location — its the GPS signal of the early fly. The Nanos protein, also found in a gradient but from back to front, is a hack: it’s only purpose is to clear away a leaky remnant of another gene, Hunchback, which isn’t supposed to be expressed yet (although Nanos may be the diminished rump of a more elaborate ancestral posterior patterning scheme). And the Torso related genes are specifically involved in ‘capping’ the front and back ends of the fly.

The main point of interest about the terminal genes like Torso is their mechanism: we sometimes talk about maternal genes as like a paint-by-number system in which Mom lays out the lines for different areas of differentiation in Baby, and then the embryo fills in the details. The terminal genes are like a perfect example of that: in the follicle, cells literally paint the vitelline membrane of the fly with different informational molecules during the construction of the egg, and then as the embryo develops, these molecules trickle across the perivitelline space (a gap between the outer membrane and embryo proper) to bind receptors and trigger regional differentiation.

It’s also a nice segue into the dorsal/ventral patterning genes, because flies do something similar there: proteins imbedded in discrete regions of the vitelline membrane diffuse to Toll receptors, where they selective activate the Dorsal protein by freeing it from the Cactus inhibitor. We go from a paint-by-number kit to a restored gradient from back to belly side of localization of free Dorsal protein to the cell nucleus. By the way, in case they were getting bored with flies, Dorsal is homologous to NF-κB in us vertebrates, using the same nuclear exclusion/inclusion mechanism, and NF-κB is a hot molecule in biomedicine and cancer research right now.

That was my hour. I closed by threatening them with talk of zygotic genes, specifically the gap genes, next week.

Also, Wednesday we’re going to try something a little different. We’ve finished chapter 5 of Carroll’s book Endless Forms Most Beautiful so they should be ready to weigh the importance of various mechanisms, so I split the class in two and told half of them to read Wray’s article on the importance of cis-regulatory mutations in evolution, and Hoekstra and Coyne’s article that argues for a more balanced emphasis. I’d love to have a fight break out in the room.

Today was the due date for the take-home exam, which meant everything started a bit late — apparently there was a flurry of last-minute printing and so students straggled in. But we at last had a quorum and I threw worms and maggots at them.

The lab today involves starting some nematode cultures so I gave them a bit of background on that. They’re small, transparent hermaphrodites that can reproduce prolifically and will be squirming about on their plates this week. They’re models for the genetic control of cell lineage and also for inductive interactions: I gave them the specific example of the development of the vulva, in which a subset of cells in close proximity to a cell called the anchor cell develop into the primary fate of forming the walls of the vulva, cells slightly further away follow a secondary fate, forming supporting cells, and cells yet further away form the hypodermis or skin of the worm. I had them make suggestions for how we could test that the anchor cell was the source of an inductive signal, and yay, they were awake enough at 8am to propose some good simple experiments like ablation (should lead to failure of the vulva to form) or translocation (should induce a vulva in a different location). I also brought up genetic experiments to make mutants in the signal gene, in the receptors, and deeper cell transduction pathways.

All those experiments work in the predicted ways, and I was able to show them an epistasis map of the pathways. Two lessons I wanted to get across were that we can genetically dissect these pathways in model organisms, and that when we do so, we often find that toolkit Sean Carrol talks about exposed. For instance, in the signal transduction pathway for the worm vulva, there are some familiar friends in there — ras and raf, kinases that we’ll see again in cancers. And of course there are big differences: mutations in ras/raf in us can lead to cancer rather than eruptions of multiple worm vulvas all over our bodies, because genes downstream differ in their specific roles.

Then we started on a little basic fly embryology: the formation of a syncytial blastoderm, experiments with ligation and pole plasm manipulation in Euscelis that led to the recognition of likely gradients of morphogens that patterned the embryos. From there, we jumped to the studies of Nusslein-Volhard and Wieschaus that plucked out the genes involved in those interactions and allowed whole new levels of genetic manipulation. As the hour was wrapping up, I gave them an overview of the five early classes of patterning genes: the maternal genes that set up the polarity of the embryo; the gap genes that read the maternal gene gradient and are expressed in wide bands; the pair-rule genes that respond to boundaries in gap gene expression and form alternating stripes; the segment polarity genes that have domains of expression within each stripe; and the selector genes that then specify unique properties on spatial collections of segments.

And that’s what we’ll be discussing in more detail over the next few weeks.

Slides used in this talk