Dynamic Development

Main Page Dynamic Development

The Foundations of Developmental Biology


From Sperm and Egg to Embryo

Genetic Regulation of Development

Organizing the Multicellular Embryo

Generating Cell Diversity

Dynamic Development at a Glance

Learning Resources

Research Resources

The Developmental Biology Journal Club

Developmental Biology Tutorial

Establishment of Spatial Patterns of Gene Expression During Early Vertebrate Development: Hox Genes

by Dr. Derrick Rancourt
Department of Medical Biochemistry
University of Calgary

We shall now continue from the discussion of homeotic genes in Drosophila and discuss parallels that have been observed in higher eukaryotes, namely vertebrates. Much of the material that I shall be discussing is new and not in the text because this particular field has advanced considerably in the past five or so years.

Take Home Lesson:

The homeotic (Hox) complex governs both the overt and non-overt segmentation that occurs in vertebrates.

Key Features of this lecture are :

1) The vertebrate homeotic complex comprises four distinct Hox gene clusters (Hox A, B, C, D) that are organized into thirteen homology (or paralogue) groups.

2) The chromosomal organization of the genes in each Hox cluster reflects its anterior-posterior expression in the body plan (spatial colinearity). Unlike in Drosophila, vertebrate Hox genes are also temporally colinear and are expressed in an anterior to posterior direction. In general, members of the same paralogue group are expressed at the same time and have the same anterior boundary of expression.

3) Homeotic genes are expressed within segmented and unsegmented structures within the body plan. Hox gene expression in some unsegmented structures arise from segmented precursors. The specification of segmented structures may be due to a specific combination of homeotic genes (or Hox Code).

4) Modern genetic methods such as targeted mutagenesis in mouse have begun to reveal the function of homeotic genes in vertebrates. In general, homeotic phenotypes observed in specific hox gene mutations are restricted to the anterior boundary of expression.

5) The genetic redundancy of the vertebrate homeotic complex may have enabled the rapid evolution of vertebrates.

Evolution of the Homeotic Complex

Homeosis was first defined by Bateson (1894) to describe natural variants of a species where certain parts of the body plan displayed morphological features typical of other regions. This idea was reiterated by the work of Lewis, who found several homeotic mutants affecting segment identity in Drosophila. After the cloning of the Drosophila homeotic complex in Drosophila, these genes were found to be transcription factors and coined to be homeotic selector genes, meaning that the fate of a given segment was selected by the expression of specific homeotic genes. Of course, it was immediately important to see if mouse and other vertebrates had a similar set of genes - especially because vertebrates only appear to be partially segmented animals.

Once this was confirmed, a mad race began in the mid eighties to clone all of the homeotic genes, which only became complete about two years ago. Now, in the mouse and human (and presumably other vertebrates including chicken and Xenopus), there are thirty nine genes that are organized on four different Hox clusters (A, B, C, D) found at four distinct chromosomal loci. Based on homologies, between themselves and their Drosophila counterparts, these genes have been organized into thirteen homology groups called paralogues.

There are several interesting points that can be made simply by comparing the Drosophila and vertebrate complexes. First, paralogue groups 9-13 all appear to derived from Abd-A through some gene duplication event, while similarly paralogue groups 2 and 3 are derived from proboscapedia. Unlike in Drosophila, where the Antennapedia and Bithorax complexes were separated in evolution, each vertebrate homeotic complex is intact. However, within each cluster, there are genes missing, which is thought to be an important feature in our evolution. (We shall return to this point.) In Amphioxus, a near vertebrate ancestor, there is only one complex that resembles a complete Hox A cluster. That means, for instance, that hoxa-8 is present this animal and that it has since disappeared. Similarly, that ancestral homeotic cluster first quadruplicated itself four times, after which individual genes were lost in evolution. As we shall discuss later, there is considerable genetic redundancy within the homeotic gene complex, and this redundancy made it relatively easy to lose specific homeotic genes in evolution without drastically affecting the body plan.

Hox Gene Expression and Colinearity

Whereas the colinearity that occurs between the homeotic complex and anterior-posterior expression was first recognized by Lewis in his genetic experiments, it was formally demonstrated in mouse using in situ hybridization. Here, members of the hoxB cluster have different anterior boundaries of expression in the CNS and prevertebra, which reflected their relative chromosomal organization. In 1989, hox genes were given names according to the order in which they were discovered. In 1997, we know that the anteriorly expressed genes reside at the 3' end of the complex and that unlike in Drosophila (whose homeotic genes are all expressed at the same time), the mouse homeotic complex is also temporally colinear, meaning that as the body plan develops in an anterior-posterior direction during gastulation, there is a sequential expression of the homeotic complex. We also know that the paralogue groups are expressed with roughly the same anterior boundaries in segmented tissues and that the next more posterior group is expressed one or two segments more posterior.

(See Fig. 14.39, Browder et al., 1991; Gilbert, 1997, Fig. 16.3; Kalthoff, 1996, Fig. 22.6; Wolpert et al., 1998, Fig. 4.8)

There are some interesting differences in the regulation between vertebrate and Drosophila homeotic genes.

  • First, each vertebrate hox cluster is the size of the Ubx gene in Drosophila. This observation has been used to argue that the regulation of the Drosophila complex is more complex. Indeed, whereas we have observed shared regulatory elements in the vertebrate genes, in Drosophila each homeotic gene appears to function more or less independantly.
  • Second unlike in Drosophila, there are no equivalent pair rule and segment polarity genes setting up segmentation for the homeotic genes. While many homologues of these genes exist, such as even skipped and paired, these homologues are often expressed after the homeotic genes are expressed. Thus, it is still a mystery how vertebrate hox genes are turned on in embryogenesis.

In vertebrates, the main evidence of segmentation is in the vertebral column. The head, the fore- and hindlimbs are later adaptations onto the early vertebrate skeleton. The earliest expression studies focused on the expression of hox genes in the prevertebra of mid-gestation embryos. In these experiments, it was realized that specific paralogue groups had anterior boundaries within prevertebra and that each vertebra had its own hox address or code. Whereas the anterior boundaries are well defined, the posterior boundaries of hox genes are not; suffice it to say that many more homeotic genes are expressed within posterior segments than in anterior segments.

Hox Gene Expression and Embryogenesis

Recall that during neurulation, the vertebral precursors, somites, condense out of axial medoderm as the neural folds emerge from Hensen's node. Whereas the somites are the earliest and most overt evidence of the segmentation at this time, there is a theory that the nervous system is segmented as well. As the neural folds are emerging and before we begin to see somites, electron microscopists have observed transient stuctures within the neural folds called somitomeres. Although these structures will disappear at somitogenesis, it is believed that the basic neural tube is segmented, although there is little overt evidence to demonstrate this. However, the existence of somitomeres helps to explain the boundaries of hox gene expression within the midgestation CNS.

However, if we look at the CNS earlier in embryogenesis, shortly after neurulation, and during somitogenesis, segmentation is obvious in the rhombencephalon, which is the CNS precursor of the hindbrain. In this structure and at this time there is a segmental periodicity to the expression of the most anteriorly expressed genes. So, each rhombomere has its own hox code as well (i.e., except R1 and R2, which do not express hox genes.

(Browder et al., 1991, Fig. 14.40; Gilbert, 1997, Fig. 16.4; Kalthoff, 1996, Figs. 22.7-22.8; Wolpert et al., 1998, Figs. 4.23-4.24)

Hox gene are not expressed anterior to the hindbrain. The forebrain and midbrain are more advanced structures in vertebrate and although there is evidence that these structures are also segmented, the homeotic gene family had nothing to do with the elaboration of these structures. An important feature of the rhombencephalon is that cranial neural crest cells delaminate from this structure as the neural tube is closing. These specialized neuroectodermal cells are a more recent adaptation of jawed vertebrates and will give rise to, or contribute to, the formation of the jaw, the cranial nerves and soft tissues of the neck and throat, including the thyroid and thymus. Anterior hox genes that are expressed in rhombencephalon are also expressed in their derivative neural crest cells. These neural crest cells will migrate to the branchial arches and give rise to these specialized structures. Each branchial arch therefore has its own hox code, and - as we shall see - mutations within the anterior genes can affect the emergence of these specialized tissues.

Whereas hox gene expression has been closely studied in cranial neural crest, little is known about the role of hox genes in trunk neural crest. Although various hox gene have been found to be expressed in soft tissues such as the lung, gut, etc., little is known about whether this expression is the result of segmentation (i.e., being derived from trunk neural crest migration) or whether this expression is de novo.

The Hox Code

The hox code model predicts that combinations of hox genes specify the development of the anterior-posterior axis. Evidence for the code comes from three lines of experimentation:

  • Comparative Anatomy indicates that the appearance of specific vertebrae such as the cervico-thoracic junction is correlated with the expression of specific hox genes. HoxC6 expression is invariantly in the cervicothoracic prevertebra in mouse, chicken, goose, Xenopus and zebrafish. Interestingly, the length of the neck is different in each species. Similarly, the positioning of the lumbral-sacral junction may be a function of members of the hox10 paralogous family.
  • Retinoic acid induced homeosis during mouse embryogenesis. Retinoic acid, a developmental steroid and teratogenenic agent, has been found to alter the axial expression of homeotic genes when administered during somitogenesis. Each hox cluster has cis elements that respond to retinoic acid and thus normal developmental regulation is perturbed. The anterior expression boundaries of several anterior hox genes were shifted posteriorly, resulting in the anteriorization of vertebrae. Similarly, posterior gene expression was shifted anteriorly, resulting in a posteriorization of fate. Similar results were observed in cranial neural crest derivatives of the hindbrain: Expression boundaries shifted anteriorly, and R2 and R3 resembled the characteristics of R4 and R5.

    (See Gilbert, 1997, Fig. 16.5; Kalthoff, 1966, Fig. 22.18; )
  • Transgenic induced homeosis in mouse. Gain of function mutations, in which a posterior gene is expressed more anteriorly, resulted in a posteriorization of anterior vertebrae. This obsevation led to the suggestion of posterior prevalence, meaning that when moved to an anterior location, a posterior gene had priority over its anterior collegues. Loss of function mutations, in which a gene is removed by targeted mutagenesis, resulted in an anteriorization of vertebrae. In this situation, the hox code of the mutant prevertebra or rhombomere would now partially resemble its near-anterior neighbour. Following the posterior prevalence model, the respecified vertebra no longer expresses the posterior hox gene, which made this segment different than its anterior neighbour.

Targeted Mutagenesis in Mice

More recently, other types of manipulations have been possible. Reporter genes such as lacZ can be integrated into a locus so that it is under the control of native elements (see Targeted Modifications in Mice). Gene swaps can be performed so that when applied to the homeotic complex, a posterior gene can be made more anterior and visa versa. Very subtle mutations can be made such as point mutations. In this case, the neo cassette is placed within an intron so that expression of the gene is not perturbed. We can also make small and large deletions within the complex to destroy regulatory elements. Although people have been able to delete an entire hox complex in ES cells, they have thus far been unable to generate these mice.

I was a postdoctoral fellow in the laboratory where gene targeting was developed. Part of our interest was to apply this new technology to understanding the homeotic complex. There were approximately 30 genes known in the complex at the time, and we undertook the insurmountable task of knocking out each and every gene in the complex. By the time I left, this work was about three quarters complete. I want to highlight a few interesting experiments that shed light on the complexity of this multigene family.

The first gene that we disrupted was hoxA3,and the phenotype that was observed was soft tissue defects in the neck and upper chest. For example, the glottis and epiglottis were aberrant, the thymus and thyroid were reduced and there were significant alterations of the aortic arteries. Because all of these structures were derived from the neural crest, which invaded the fourth branchial arch, these results were the first to demonstrate that hox gene expression was important for specifying the fate of neural crest cells.

This result has been confirmed with mutations in other anterior genes such as hoxB2 ,which is expressed up to the R1 boundary. Interestingly in this mutation, bones of the jaw, which arise from neural crest in the second branchial arch, are duplicated posteriorly as a result of an anteriorizing homeotic mutation. This is "heady" stuff here, but the interesting part of this mutation, is that the jaw structure of this mutant resembles that of reptiles and thus this hoxB2 mutation is considered to be atavistic - meaning that it is found in ancestral animals.

An interesting observation that was made following the A3 mutation was that mutations in the paralogue D3 did not result in a soft tissue defect but resulted in a skeletal defect, suggesting that - whereas paralogues appeared to have similar expression boundaries - their roles were distinct in different tissues. In this case, the first cervical vertebra was fused to the occipital bone, which was considered to be an anteriorizing homeotic mutation.

At this time, of course, we wondered what the A3, D3 double mutant might look at. What was emerging in the gene targeting field was the idea of genetic redundancy; i.e., that phenotypes in mice could be made less severe because they were masked by the function of related genes. Genetic redundancy was confirmed when A3, D3 double mutants were generated. Interestingly, one or two mutant copies of the D3 gene were found to exacerbate the A3 soft tissue phenotype, whereas one or two mutant copies of the A3 were found to exacerbate the D3 skeletal phenotype. Thus, it appears that - whereas both hox genes have a primary responsibility in soft tissue or skeletal development - they appear to be able to contribute independently to each function.

Genetic redundancy was not restricted to the paralogous member, however, because in experiments that I performed, we observed similar genetic interactions between neighbouring hox genes. Now we know that a complex web of interactions govern the homeotic complex and that neighbouring paralogues and even next to neighbouring paralogues can augment each others' phenotypes just like in the case of A3, D3.

As a result, a considerable amount of genetics will still be required to understand the complexity of the homeotic complex.

Genetic Redundancy of the Homeotic Complex Permits Evolution of the Body Plan

With the exception of some of the more anterior genes in the complex, mutations in many hox genes resulted in only minor phenotypes that did not affect morphology greatly. Often, in mutations affecting vertebral structure, only one side of the animal was affected, suggesting that there is some plasticity in the response hox code. Hox genes have long thought to play an important role in limb development. However, when limb genes such as A11 or D11 were made, the phenotypes were hardly detectable. On the other hand, once two mutations were put together, the effects were dramatic. For example, the absence of both A11 and D11 results in a severe and life threatening reduction in the radius and ulna. What this result tells us is that the redundancy of the Hox complex permits some flexibility in the response to mutational change.

Redundancy permits a rich potential to allow for evolutionary alterations of the body plan, through the following mutational changes:

  • variations in the number of homeotic genes, by deletion and or duplication;
  • increases in the number of hox complexes through whole complex duplication events;
  • mutations affecting the timing, position or level of homeotic gene activation, which may be most relevant to generate small adptive changes;
  • alterations in the regulatory interactions between Hox proteins and their targets, through muations of the coding sequenc of Hox genes.

Recently, heterochronic mutations have been made within the hox gene complex by changing the position of a given gene within the complex. When hoxD11 is targeted into D13, a significant limb phenotype results - not from the inactivation of hoxD13 itself but from the inappropriate activation of D11 at the wrong time.

Digging Deeper:

Recent Literature

Mann, R.S. 1997. Why are Hox genes clustered? BioEssays 19: 661-664.

Morrison, A., Ariza-McNaughton, L., Gould, A., Featherstone, M. and Krumlauf, R. 1997. HOXD4 and regulation of the group 4 paralog genes. Development 124: 3135-3146.


Bateson, W. 1894. Materials for the Study of Variation. Macmillan. London.

Browder, L.W., Erickson, C.A. and Jeffery, W.R. 1991. Developmental Biology. Third edition. Saunders College Pub. Philadelphia.

Gilbert, S.F. 1997. Developmental Biology. Fifth Edition. Sinauer. Sunderland, MA.

Kalthoff, K. 1996. Analysis of Biological Development. McGraw-Hill. New York.

Wolpert, L., Beddington, R., Brockes, J., Jessell, T., Lawrence, P. and Meyerowitz, E. 1998. Principles of Development. Current Biology. London.

Dynamic Development at a Glance
Main Page Dynamic Development

Dynamic Development is a Virtual Embryo learning resource

This material may be reproduced for educational purposes only provided credit is given to the original source.
Leon Browder & Laurie Iten (Ed.) Dynamic Development
Last revised Monday, July 20, 1998