Hox  genes  in  evolution  and  development

A great variety of species has arisen over the course of evolution by acquiring distinct morphological traits that become specified during embryonic development. Although the coupling of evolution and development (Evo-Devo) has long intrigued biologists, it was only with the advent of molecular genetics that this discipline truly emerged with the finding that key developmental molecules are conserved throughout a variety of different phylogenetic clades. These discoveries first started with the realization that Hox genes were present in distinct animal phyla that undergo very different developmental programs (Moczek et al., 2015).

Hox genes, formerly termed homeotic selector genes, were first described in the fruit fly (Drosophila melanogaster) by Ed Lewis as genes whose misexpression resulted in the change of segment identity of a segment into another. These phenotypes, termed homeotic transformations, revealed the crucial role of homeotic genes in fruit fly patterning.

Furthermore, it was proposed that Hox genes could have been implicated in the evolution of Drosophila from its four-winged ancestor (Lewis, 1978).

All homeotic genes in the fruit fly were found to contain a DNA sequence named homeobox. This sequence is 183 base pair-long and codes for a helix-turn-helix 61 amino acid motif (homeodomain) with DNA binding ability. Subsequently, homeobox-containing genes were discovered in the genome of many other animal species, including vertebrates (McGinnis et al., 1984). These discoveries were followed by the realization that many important players in embryonic development were conserved throughout evolution. With this knowledge, new interdisciplinary approaches have arisen that made it possible to tackle biological questions in a variety of different fields.

1.1.1 Hox gene clustering

Hox genes code for transcription factors that have been found in most metazoans including cnidarians and all bilaterians (Lemons and McGinnis, 2006). An ancestral ProtoHox cluster containing few genes is thought to have arisen before the split between cnidarians and bilaterians over 1000 million years ago. After analyzing Hox genes in cnidarian species, it was proposed that this ProtoHox cluster contained only two anteriorly expressed genes. The cluster would have been duplicated, originating identical Hox and

ParaHox clusters, which subsequently diverged and expanded by tandem duplications in cis (Garcia-Fernandez, 2005). The sequencing of Hox genes in different bilaterian species has motivated the hypothesis that a minimum of seven genes would have been present in the ancestral bilaterian cluster (de Rosa et al., 1999). However, this initial cluster has evolved differently in different species acquiring distinct organizations and Hox gene number.

In the fruit fly Drosophila melanogaster, where Hox genes were first identified, a set of eight genes located on chromosome 3 is split in two complexes separated by eight megabases: Antennapedia (ANT-C) and Bithorax (BX-C) (Lewis, 1978). Later, vertebrate Hox genes were also found to be organized in a cluster, which led to the erroneous assumption that the presence of Hox genes in a species was necessarily associated with a cluster-like organization (Duboule, 2007). Furthermore, clustered Hox genes were thought to be essential for the correct spatial expression of Hox genes and by consequence to be under a strong functional constraint (Lewis, 1978). However, the increasing data on Hox gene repertoire and organization of different animals has shown that this is not always the case. For example, Nematostella vectensis and Eleutheria dichotoma, two cnidarians, have shown to have so-called atomized Hox clusters (Duboule, 2007) that are very unorganized, with Hox genes isolated or in small groups scattered in the genome (Kamm et al., 2006).

Interestingly, species closely related to Drosophila melanogaster have their two Hox complexes split at a different location. More basal insect and diptera species, however, have an intact cluster indicating this to be the insect ancestral condition (Ferrier and Minguillon, 2003). However, although this cluster is present at a single locus in the insect genome, the intergenic regions are large and the transcription orientation is not always the same (Duboule, 2007).

In what concerns echinoderms, the sea urchin presents a very interesting Hox gene organization. It would appear that Hox genes, although organized in one single locus, have rearranged their order in the cluster in comparison with the ancestral condition (Cameron et al., 2006). Echinoderms are the sister group of hemichordates but are less diverged and show more similarities with chordates. The analysis of Hox gene organization in two species of this phylum has revealed a cluster composed of 12 genes in the correct order. The 10 first genes have the same transcriptional orientation as the chordate counterparts, while the two more posterior genes are inverted (Freeman et al., 2012).

Echinoderms and Hemichordates compose together the group of Ambulacraria which is a sister group of Chordates. Within the Chordate phylum Cephalocordates (Amphioxus) represent the most basal subphylum (Putnam et al., 2008). The Amphioxus Hox cluster contains 15 Hox genes in the same locus displaying the same transcriptional orientation and has not suffered any rearrangements. While the recently discovered 15^th Hox gene has only been discovered in the Amphioxus genome (Putnam et al., 2008), Hox14 paralogs have been

found in vertebrate species such as the lamprey, cartilageneous fishes or coelacanth. This gene is thus unlikely to result from an amphioxus-specific duplication. However, Hox14 appears to have a distinct function from other Hox genes and thus to be under a more relaxed selective pressure, which could explain why it is missing in most species of the vertebrate lineage (Kuraku et al., 2008).

Urochordates, the sister group of vertebrates, have a dramatically different Hox gene organization. Ascidians, for example, have been shown to have nine Hox genes distributed in two different chromosomes with a highly rearranged organization and large intergenic regions containing unrelated genes (Ikuta et al., 2004). Another Urochordate, Oikopleura dioica presents an even greater level of disorganization (Seo et al., 2004).

Early in the vertebrate evolution, two events of genome duplication occurred that originated four Hox clusters. As a consequence, over the course of vertebrate evolution, several events of gene loss occurred and, to date, no extant vertebrate has been shown to possess a complete Hox cluster (McClintock et al., 2001; Wagner, 1998).

Within the vertebrate clade, there are two superclasses: Agnatha (jawless vertebrates like lampreys and hagfish) and Gnathostomata (jawed vertebrates). Agnathan Hox clusters are large and have a high repeat content, superior to the percentage of repeats in the genome (Mehta et al., 2013). The Japanese lamprey has six clusters instead of four most likely due to segmental duplications (Mehta et al., 2013; Smith and Keinath, 2015).

In Gnathostomes most lineages have four Hox clusters, except for some fish groups that suffered additional large-scale duplications resulting in a higher number of clusters (seven in zebrafish, for example). In comparison to their Agnathan counterparts, gnathostome Hox clusters tend to be more compact and lower in repeat content (Mehta et al., 2013).

In particular, all mammalian species analyzed so far have 39 Hox genes organized in four clusters, HoxA, HoxB, HoxC and HoxD that are located in four different chromosomes.

All Hox genes are ordered and have the same transcriptional orientation within each cluster of about 100 kilobases in size. Based on homology, thirteen paralogous groups have been identified although only groups 4, 9 and 13 are represented in all clusters (Lemons and McGinnis, 2006).

The increasing genomic information supports a model whereby an ancestral bilaterian cluster would have suffered extensive changes over the course of evolution to the point of Hox genes being scattered in the genome surrounded by unrelated non-Hox genes in an atomized conformation. It is therefore surprising to find gnathostomes as the only bilaterians to have countered this trend by consolidating their clusters. The increase in cluster and Hox gene number would be expected to confer a higher flexibility permitting genes to be rearranged without resulting in adverse effects. This unique consolidation observed in gnathostomes is therefore likely to have presented an important advantage to the lineage. The

implementation of global long-range enhancers that require the close proximity of genes for a coordinate regulation to take place has been proposed as an explanation (Duboule, 2007).

1.1.2 Collinearity

One of the most fascinating aspects of Hox gene expression first reported in Drosophila melanogaster by Ed Lewis was that the position of Hox genes in the cluster reflected the location of their gene expression domains in the developing embryo (Lewis, 1978). This phenomenon, termed spatial collinearity, was subsequently found to be conserved in mice as well (Duboule and Dollé, 1989; Graham et al.). Therefore, genes located at the 3’

end of the cluster such as Hox1 genes are expressed in more anterior domains of the embryo, while genes located in the 5’ end of the cluster are expressed in the embryonic tail. Spatial collinearity has since been found both in vertebrate and invertebrate species regardless of the organization level of the cluster (Monteiro and Ferrier, 2006).

In species that have a progressive axis formation and segmentation, such as vertebrates and short-germ band insects, 3’ Hox genes are not only expressed more anteriorly but they are also activated earlier in development, whereas 5’ located genes start to be expressed later (Kmita and Duboule, 2003). This phenomenon, termed temporal collinearity appears to require a cluster-like Hox gene organization (Tschopp et al., 2009). Details on the regulation mechanisms behind collinearity are detailed in section 1.3.3.

1.1.3 The Hox code and posterior prevalence

The collinear activation of Hox genes in nested overlapping domains has generated the hypothesis that the combination of Hox genes in a given anterior-posterior (AP) region of the embryo would determine its fate – the Hox code (Kessel and Gruss, 1991). Loss and gain of function experiments are however inconsistent with a combinatorial activity of all Hox transcription factors expressed in a tissue. For example, Hox gene overexpression doesn’t result in homeotic transformations where more posterior genes were expressed (Duboule and Morata, 1994). These observations led to the hypothesis that Hox genes act hierarchically, with more posterior genes overruling anterior gene activity. This system is referred to as

“posterior prevalence”. In some cases, however, a combined activity of different paralogous groups has been shown to take place, for example in mouse rib cage patterning (McIntyre et al., 2007).