The sunday evening symposium had a grand title: "Homeoboxes—twenty years on, how do they explain development?" Unfortunately, no one even attempted to answer the question in the subtitle. It was a good session, but it was really all nitty-gritty down-and-dirty analyses of epistasis in the Hox genes. Working science. Very little big-picture speculation.

We have a good idea about how Hox gene expression patterns are set up in Drosophila (thank you, Ed Lewis, and fly labs all over the world), and all three talks in this symposium were entirely about vertebrate Hox genes and how they are regulated. Cecilia Moens gave away the story for the whole evening in her introduction. She said that there were three big points to make:

• Hox regulation is different in flies and vertebrates.
• Hox genes are both auto- and cross-regulatory.
• Retinoic acid plays an important role in vertebrate Hox gene patterning.

Fly Hox genes are activated by a set of genetic players we know well: gap genes, pair-rule genes, and segment polarity genes. The weird thing is that although vertebrates have the same colinear arrangement of Hox genes on the chromosome and have retained the same anterior-posterior association of Hox gene expression with regions of the body, there are huge differences in how the spatial pattern is set up and maintained. In vertebrates, there is a gradient of retinoic acid that seems to be an important (but not the only) regulator of where Hox genes will be expressed. Hox genes are also auto-regulatory. That is, they amplify their own expression in any segment where a low level of expression is triggered. They are also cross-regulatory, which means one Hox gene works to influence the expression of another, often in an inhibitory way, which helps to set them up in tightly bounded, exclusive zones.

Moens talk then plunged into the details. She's looking at how stripes of Hox gene expression are turned on in the zebrafish hindbrain (and oooh, but she had gorgeous multicolored in situs of precisely, sharply delimited rainbow stripes in the hindbrain—sometimes, you do see straight lines in biology, and they are startling to see in a curvy, complicated, organic structure like the brain). And then it got complicated. She described experiments and analyses in which she modified different components of the Hox system by mutation or pharmacology, and looked at how the expression of other components then changed. And she was illustrating it graphically—my notes are full of bars and wedges with arrows between them, that are rather difficult to express in text. The capsule summary: she's looking at genes in rhombomere 4, 5, 6, and 7 of the hindbrain. Retinoic acid is present in a theorized gradient, low at the front and high at the back. Low retinoic acid levels activate Hox1 and FGF in R4. Intermediate levels activate Hox3 in R5 and R6 via the genes vhnf1 and valentino/kreisler. In R7, valentino is suppressed by an unidentified factor that is activated by Hox4, which is regulated by high retinoic acid levels.

Summarized that way, it all sounds like minor details, but trust me—it was all real science, that stuff where an investigator carefully lays out every step of her experiments and shows you both the data and the logic of her interpretations. I had a good wonky time listening to it.

The second talk was by Rob Krumlauf, and it was more of the same: detailed analyses of the regulation of vertebrate Hox genes in the hindbrain. The executive summary of this talk is that the Hox B1 gene contains distinct regulatory regions that mediate the gene's response to retinoic acid, it's autoregulatory properties, and its interaction with other Hox genes. When cells are transplanted from one Hox domain to another, he can induce interconversion to a new regional identity if the clumps of cells are small enough; large clumps set up their own local domain that preserves the identity of the original location.

One really cool aspect of Krumlauf's work is that he uses both mouse and chick models. Mice are powerful organisms because we have so many genetic tools to manipulate them; unfortunately, they're mammals with internal development, which makes them a pain to work with in those inaccessible fetal stages. The chick, on the other hand, is a big embryo, easy to manipulate, and it's in an egg which you can culture and tinker with freely at any point in development...but the genetical tools in birds are relatively primitive. The solution? Make mutants in mice, and transplant the cells to chick embryos. One of those amazing indicators of the unity of life is the fact that mammalian cells and bird cells respond to the same developmental signals and use the same molecules to build their brains, and you can maximize the strengths of both systems by bouncing freely between the two.

The last talk of the evening did approach the grand promise of the session's title. It was by Mario Capecchi, and was called "Hox genes: from body plan to neuropsychiatric disorders." Capecchi essentially gave us three short, easily digestible vignettes on different aspects of Hox gene function.

The first story addressed another difference between flies and vertebrates. In flies, we have lots of clear-cut homeotic transformations: we can turn antennae into legs, or wings into halteres. We don't see anything so patent in us vertebrates. The reason is that vertebrates have at least four banks of co-expressed Hox genes, which means that there is a lot of redundancy and combinatorial encoding of segment identity: knock out one gene, and there's another that can at least partially cover its loss. The obvious experiment to test that explanation is to knock out all of the paralogs in a region, and see what kind of homeotic transformation you get. Capecchi has generated two triple mutants. One deletes HoxA10, HoxC10, and HoxD10 (there is no HoxB10), and the other gets rid of HoxA11, HoxC11, and HoxD11 (again, no HoxB11). These genes affect the caudal part of the body. Losing all of the Hox10s does a nifty thing: all of the vertebrae in the lumbar and sacral regions are transformed into thoracic vertebrae, and develop ribs! Hox10 is a suppressor of thoracic identity. Taking out all of the Hox11s does something slightly different. In this case, sacral structures don't develop; the stubby little 'mini-ribs' that characterize vertebrae in that region don't appear. Hox11 is a partial suppressor of Hox10. This fits with the way Ed Lewis portrayed Hox function in the fly, as largely a series of negative interactions that suppress and modify a common ground state. In vertebrates, it looks like the ground state is to make ribby thoracic segments everywhere.

Capecchi's second vignette was to show a later function of the Hox genes. Pleiotropy is a universal feature of genes—they do more than one thing, and affect more than one process. In this case, he is arguing that the Hox genes also assist in neural development by maintaining the register between the input and output of the central nervous system. Specific Hox genes are expressed in the brain, and that same Hox gene is also expressed in the peripheral tissues that the neurons in that region will innervate. For example, the HoxB1 gene is expressed in both nuclei of the facial nerve, and in muscles of the face. A HoxB1 knockout has the effect of inducing facial paralysis, since the neurons don't form or can't find their way to appropriate regions of the face. He illustrated this with pictures that surprised me a bit. Did you know that mice have facial expressions? He showed pairs of photographs of mice that were startled or irritated alongside HoxB1 knockout mice under the same conditions, who always just looked phlegmatic.

His third story pursued pleiotropy further. What do Hox genes do after early embryogenesis? Some are still actively expressed in the adult brain, in characteristic locations. HoxB8 is found in the adult brain in the olfactory bulb, the caudate and putamen, the hippocampus, orbitofrontal cortex, the anterior cingulate, and the brainstem. Mice with HoxB8 deletions do something pathological: they are obsessive groomers. They constantly scrub their little faces and stroke their pelts, and if other mice are kept with them, the HoxB8^- mice will constantly groom them. They groom so incessantly that they develop great bald spots and skin lesions—and other poor mice housed with them also lose their hair. The mice have obsessive-compulsive disorder and trichotelomania! Interestingly, the pathways thought to be affected in human OCD are the caudate-putamen, orbitofrontal cortex, and anterior cingulate—just the areas that also express HoxB8.

There are two lessons to take from this. One is that while there is lots of redundancy and overlap in body plan functions such that it is difficult to generate obvious homeotic transformations in vertebrates, secondary functions, like HoxB8's role in the CNS, are more phenotypically susceptible. The second lesson is a bit dubious, and plays into an unfortunate bias in genetic research: that we can explain complex properties of organisms by single gene effects. We might be able to explain a defect as a consequence of a change in a single gene, but we have to be careful not to fall into the trap that the gene is causal to the normal behavior (I don't think Capecchi has made that mistake, but I can imagine how newspapers would handle this story). Grooming behavior is the result of a multitude of genetic and epigenetic factors. Damaging one of those factors could cause aberrant behaviors.

Capecchi is currently taking advantage of all that human genetic information they have out there in Utah to look for polymorphisms in HoxB8 in humans, and to see how they correlate with OCD. He made the point, though, that this is such a common syndrome, affecting roughly 3% of the population, that it was relatively easy to find samples; he estimated that there were about 20 people in his audience that night who would be good subjects for his research.