Head to Tail: Hox Genes, Retinoic Acid and Rhombomeres

Isotretinoin, a form of retinoic acid made from vitamin A, went on sale in the United States in 1982 as a treatment for severe acne. By 1985, Edward Lammer and colleagues had described a pattern of defects in children exposed to it early in pregnancy: small or missing ears, an altered face and jaw, heart defects, and abnormalities of the brain.

They called it retinoic acid embryopathy. A pill taken for the skin had reached the tissue that decides which part of the head becomes the jaw and which becomes the face. To see how, you need to know how the neural tube gets its addresses, and what retinoic acid has to do with a family of genes called Hox.

The idea: the tube gets an address before it gets neurons

The neural tube is the hollow rod that forms when the edges of the neural plate fold up and fuse. It becomes the brain and spinal cord. Its front end soon swells into three vesicles: the forebrain (prosencephalon), the midbrain (mesencephalon) and the hindbrain (rhombencephalon). The forebrain then divides into the telencephalon, which makes the cerebral hemispheres, and the diencephalon, which makes the thalamus and its neighbours and pushes out the optic vesicles that become the retinas. The hindbrain divides into the metencephalon (pons and cerebellum) and the myelencephalon (medulla).

The hindbrain also shows a row of small bulges called rhombomeres, numbered r1 to r7 from front to back, with a less distinct r8 behind. Repeated units like these are called neuromeres. Early on, the cells of one rhombomere look much like those of the next. Yet r2 will make the motor neurons that work the jaw, and r4 the ones that move the face. So the question becomes: how is the nervous system specified within neuromeres?

Think of a new street where every house is built to the same plan. Before anyone moves in, each house needs a number. The embryo’s answer rests on three basic principles that will come back in the lessons on the tube’s other axes:

1. Particular genes are switched on at particular places and particular times.
2. Signals that spread from a source and oppose one another form graded concentrations, and those gradients set the pattern.
3. The signals restrict where certain transcription factors are made, and those factors then regulate the genes that give each region its character.

In the hindbrain, the signal is retinoic acid and the transcription factors are the Hox proteins.

Hox genes: order on the chromosome, order along the body

The story starts in the fruit fly. A homeotic mutation turns one body part into the likeness of another: legs where antennae should be, or a second pair of wings in place of the small balancing organs called halteres. Edward Lewis spent decades on one cluster of such genes, the bithorax complex. In 1978 he proposed that its genes lie along the chromosome in the same order as the body segments they control, from front to back.

In 1984 two groups found what these genes share. The laboratory of Walter Gehring in Basel (William McGinnis, Michael Levine, Ernst Hafen, Atsushi Kuroiwa and Gehring) and the laboratory of Matthew Scott at Indiana (Scott and Amy Weiner) each found a stretch of about 180 base pairs common to several fly homeotic genes. It was named the homeobox. It encodes the homeodomain, about 60 amino acids that later work showed bind DNA, with one helix lying in the major groove. The Basel group soon detected related sequences in vertebrate DNA. Body-plan genes belong to a shared animal toolkit (see that lesson).

Hox genes are one branch of the much larger homeobox family, and they come in clusters. The fly has eight, in two complexes on one chromosome. Mice and humans have 39, in four clusters called Hoxa, Hoxb, Hoxc and Hoxd, each on a different chromosome. Genes that sit at the same position in different clusters, such as Hoxa3, Hoxb3 and Hoxd3, form a paralog group. There are 13 groups, and no cluster has all 13. Paralogs have similar sequences and often similar expression, so they can stand in for one another, and knocking out a single Hox gene often gives a milder defect than you might expect.

The clusters obey the rule Lewis saw, called colinearity:

• Spatial colinearity. Genes at the 3′ end of a cluster (group 1) have the most anterior front edges, reaching into the hindbrain. Genes farther toward the 5′ end have edges farther back, and group 13 is expressed near the tail.
• Temporal co-linearity. In mouse and chick, 3′ genes are also switched on first and 5′ genes later, as the body axis grows backward. This is a vertebrate observation. In the fly the timing rule is weak or absent.

The somites, blocks of mesoderm beside the tube that form one pair at a time (roughly every 90 minutes in chick), show temporal colinearity too, but that story belongs to them.

In the hindbrain the Hox genes with sharp front edges are mainly those of paralog groups 1 to 4, and their edges sit at rhombomere boundaries.

Retinoic acid: the vitamin that sets the back of the code

Retinoic acid is made from retinol (vitamin A) in two oxidations, through retinaldehyde. In the early embryo the main enzyme for the second step is Raldh2, made in the mesoderm beside the future hindbrain and spinal cord. A family of enzymes called Cyp26 oxidizes retinoic acid to inactive products, and Cyp26a1 is strongly expressed toward the front of the neural plate. Picture a bath with the tap at one end and the drain at the other. Supply behind and destruction in front leave retinoic acid high toward the back of the hindbrain and low toward the front.

Retinoic acid is small and oily, so it crosses membranes and reaches the nucleus. There it binds the retinoic acid receptor (RAR), which works as a pair with the retinoid X receptor (RXR) on short DNA sequences called retinoic acid response elements (RAREs). Both are nuclear receptors, ligand-activated transcription factors of the kind covered in that lesson. Hoxa1 and Hoxb1 carry RAREs, and in cell culture 3′ Hox genes respond to lower doses of retinoic acid than 5′ genes. The gradient and the cluster order fit together: where the signal is lower, fewer and more 3′ Hox genes come on.

Change the signal and the code moves. In 1992 Heather Marshall, Robb Krumlauf, Andrew Lumsden and colleagues reported that exposing mouse embryos to retinoic acid at an early stage changed the hindbrain Hox domains and transformed r2 and r3 toward the identity of r4 and r5. Removing the drain does something similar. Mouse embryos lacking Cyp26a1 showed a rostral hindbrain that looked partly transformed toward a more posterior character (Abu-Abed et al., 2001). In mouse embryos lacking Raldh2, the posterior hindbrain is poorly patterned.

A teratogen is an agent that causes birth defects when an embryo or fetus is exposed to it. Retinoids qualify because the early embryo uses this very signal to set its axis. Extra retinoic acid in the first weeks shifts the Hox code of the hindbrain and disturbs the neural crest cells that leave it to build much of the face, which fits the ears, jaw and heart in Lammer’s description. Too little vitamin A harms the embryo as well. The dose has to be right in both directions.

Rhombomeres: hindbrain rhombomere identity is determined by the Hox code

In 1989 Andrew Lumsden and Roger Keynes mapped the motor neurons of the cranial nerves in chick embryos. Branchiomotor neurons, the motor neurons for muscles of the jaw, face and throat, arose in pairs of rhombomeres: trigeminal (nerve V, chewing) from r2 and r3, facial (nerve VII, facial expression) from r4 and r5, glossopharyngeal (nerve IX, swallowing) from r6 and r7. Each nerve left from the even-numbered member of its pair. In 1990, with Scott Fraser, they injected single cells with a tracer. Clones labelled before boundaries appeared could span two rhombomeres; clones labelled after did not. The rhombomeres are compartments. One reason cells do not mix is contact repulsion: in fish and frog embryos, r3 and r5 make the receptor EphA4 while their neighbours make ephrin-B ligands, and cells sort where the two meet.

The same year, David Wilkinson, Krumlauf and colleagues reported that mouse Hox genes have expression domains whose front edges fall at rhombomere boundaries, often two rhombomeres apart. Later in development Hoxb1 is strong in r4 alone. Each rhombomere carries its own combination of Hox proteins, matched to the nerve it builds.

What about r1, which makes the cerebellum? Hox expression stops at r1, kind of. Normally no Hox gene is expressed there, and the front edge of Hoxa2 sits at the r1/r2 boundary. But that edge is held, not merely inherited. In chick, a signal from the boundary between midbrain and hindbrain keeps Hoxa2 out of r1, and changing that signal changes where Hoxa2 is found. That signalling centre has its own lesson.

What happens if we knock out Hoxb1?

A knockout is a mouse line in which one gene has been disrupted in every cell. In 1996 two groups reported Hoxb1 knockouts: Michele Studer, Lumsden, Laila Ariza-McNaughton, Allan Bradley and Krumlauf in Nature, and J. M. Goddard, Mireille Rossel, Nancy Manley and Mario Capecchi in Development. In a normal mouse, facial branchiomotor neurons are born in r4 and migrate backward to r6, where they form the facial motor nucleus. In both mutants the facial branchiomotor neurons failed to form or to migrate normally. Studer and colleagues found that r4 lost parts of its identity and that its motor neurons behaved like trigeminal neurons of r2 and r3, a conversion toward trigeminal. Goddard and colleagues reported that the motor nucleus of nerve VII failed to form.

Design problem

You knock out Hoxb1 in a mouse. Predict what happens to r4, to its motor neurons and to the face. Then answer a student’s question: in the mutant, r4 is still there as a segment. If Hoxb1 is gone, why does r4 not simply vanish, and what happens to its motor neurons?