Figure 1. Proposed model for the evolution of the homeobox containing gene clusters in humans. (a) The Hox complexes along with Evx, Mox and possibly Dlx make up the Extended Hox group, which are linked to the EHGbox and NKL gene groups to form a supercluster. (b) The Hox gene family is comprised of four clusters (A–D) of 13 paralog groups located on different chromosomes. Open squares indicate a gene loss. All Hox genes have the same transcriptional orientation, which is indicated by the horizontal arrow. As postulated by Pollard and Holland (2000), the evolution of the homeobox gene clusters presumably originated with a single ancestral homeobox gene (shown in black at the bottom). Tandem gene duplications, coupled with whole-chromosome (most probably whole-genome) duplications and occasional gene loss, resulted in the organization of genes observed in present-day humans. Early in evolution, a duplication in the ProtoHox cluster gave rise to the unlinked ParaHox cluster, which comprises the Cdx, Xlox and Gsx genes. (Adapted with permission from Patel and Prince (2000).)

Figure 2. Representative Hox expression patterns, diagram of the evolutionarily conserved Hox complexes, and their chromosomal arrangement. (a) A Drosophila embryo with rostral to the left. (b) Hox complexes of Drosophila and human. (c) Expression patterns of Hox paralog groups 1, 4, 7 and 13 in the central nervous system. The expression of paralog group 13 does not appear in the mammalian embryo as shown, as its expression is too caudally restricted. The collinear arrangement of the Hox genes in the complex is indicated relative to spatial expression (rostral versus caudal), timing of expression (early versus late) and retinoic acid (RA) sensitivity (high versus low). The division between Hox genes that have anterior borders of expression in the hindbrain or in the spinal cord is also indicated. The transcriptional orientation of Hox genes is identical and is indicated with a bold arrow below the complexes.

Figure 3. Two different models by which cofactors (CFs) could affect the function of Hox proteins. In the examples shown, the cofactors are required for activation, but similar models can also be imagined for cofactors that repress. (a) In the ‘activity regulation’ model, Hox is shown to bind cooperatively to DNA via the homeodomain (HD) in the absence of the cofactor, yet its activity is ‘masked’ by the masking factor. The interaction with the cofactor unmasks one of the interaction surfaces (Act) on the Hox protein, converting it into a transcriptionally active form. (b) In the ‘binding regulation’ model, the Hox protein is not bound to DNA in the absence of cofactor. Association of the Hox–cofactor pair permits specificDNA binding via the HD and subsequent transcriptional activation.

Figure 4. HoxB1–Pbx1–DNA crystal structure. Ribbon diagram of the Pbx1 (blue) and Hoxb1 (red) proteins bound to DNA (Piper et al., 1999). The 20 disordered residues between the Hoxb1 N-terminal pentapeptide (Tyr-Pro-Trp-Met-Lys) and the homeodomain are indicated with a dotted line. Numbering (residues 1–60) of the homeodomain is indicated for each protein. The three-residue insertion in the Pbx1 homeodomain is numbered 23a-b-c. Residues positioned N-terminal to the Hoxb1 homeodomain are indicated by negative numbers, the more negative numbers indicate the more N-terminal residues. Figure redrawn with permission from Lufkin (2000).