Thousands of papers have been written about Hox genes over the last twenty years, and cross-species comparisons using this particular family of homeodomain transcription factors have provided the critical momentum sparking the recent resurgence of the field of evolutionary developmental biology. Three remarkable phenomena that have fueled enormous interest in Hox genes are the Hox code, Hox clusters, and Hox colinearity.
In a phylogenetically diverse range of animals, a conserved “Hox code” is partially responsible for patterning the primary body axis. The term Hox code was first applied to the segmentally-restricted expression of Hox genes in the branchial system of the developing mouse 
. However, extensive similarity among Hox expression patterns in a wide range of taxa soon led to the recognition that a Hox code might be a fundamental developmental mechanism of animals 
. In all bilaterian animals that have been studied, multiple Hox genes are found in their genomes and, over the course of development, different regions along the primary body axis come to express different Hox genes or different combinations of Hox genes. Appropriate Hox expression is required to confer the appropriate regional identity upon these Hox-expressing body regions — ergo
, a Hox code. Furthermore, comparable body regions are patterned by orthologous Hox genes in distantly related taxa, so a similar Hox code appears to be widely conserved. However, this does not mean that the “Hox code” is static over evolutionary time. Hox expression patterns can vary substantially with respect to how much overlap exists between expression domains, what fraction of the primary body axis is accounted for by Hox expression, the precise axial order of different Hox orthologs, the degree of dorsal-ventral asymmetry in Hox expression, and the germ layer in which Hox genes are expressed. With regards to this last point, while Hox genes are generally regarded as exhibiting ectodermal and mesodermal expression, they are also expressed in endoderm 
In addition, across a range of bilaterian animals, Hox genes are located in conserved genomic clusters. The relative genomic organization of orthologous Hox genes is well conserved among select Ecdysozoa such as Anopheles
, and Tribolium 
; Lophotrochozoa, such as Lineus 
; and Deuterstomia, such as vertebrates and Branchiostoma 
. The origin of Hox clusters is not especially remarkable, since the Hox clusters would have arisen as a natural outgrowth of the process of tandem gene duplication. However, the persistence of Hox clusters over hundreds of millions of years in diverse metazoan lineages suggests that strong stabilizing selection must be operating.
One explanation for the conservation of genomic organization is that the proper regulation of these genes may depend upon their close physical linkage (reviewed in 
. However, in diverse bilaterian taxa (for example, Ciona intestinalis
, Caenorhabditis elegans
, Drosophila melanogaster
, D. pseudoobscura
, D. repleta
, D. virilis
, Oikopleura dioica
, Schistosoma mansoni
, and Strongylocentrotus purpuratus
), the Hox cluster has experienced breaks, undergone extensive rearrangements, or even degenerated to the point where a cluster cannot be recognized or identified 
. The degeneration of the Hox cluster does not necessarily imply that the Hox code has been abandoned, as Hox genes may continue to specify the same axial territories even after a Hox cluster has undergone extensive rearrangements. For example, it appears that all insects employ the same Hox code, but some insects have intact Hox clusters (for example, grasshopper), while others have partially degraded Hox clusters (for example, fruit flies).
It has also been observed in a phylogenetically widespread range of taxa that the relative spatial and/or temporal expression of Hox genes is correlated with their relative position within Hox clusters. This correspondence between gene expression and cluster organization has been termed colinearity 
. The existence of colinearity implies that linkage impacts gene regulation 
. However, Hox colinearity is not universal 
, and no single mechanism has been identified that can explain Hox colinearity or the persistence of Hox clusters in diverse metazoan lineages 
. Rather, it seems that several different regulatory mechanisms may contribute to the stability of Hox clusters. For example, both higher-order chromatin structure and local cis
-regulatory elements may result in coordinated regulation of neighboring Hox loci 
. Furthermore, in some taxa, most notably Drosophila
, Hox linkage does not appear to be required for appropriate Hox expression 
. The general correspondence between the genomic organization and spatial expression of Hox genes in Drosophila
may be attributable to phylogenetic inertia.
Over evolutionary time, the functional diversification of Hox genes has clearly contributed to the diversification of animal body plans 
. For this reason, understanding the origin and early evolution of Hox genes could prove critical to understanding the metazoan radiation. A Hox cluster consisting of seven genes evolved prior to the divergence of protostomes and deuterostomes 
and, as both insects and vertebrates utilize Hox genes to pattern a portion of their primary body axes, the Hox code can be said to predate the diversification of crown bilaterians. The phylum Cnidaria can provide unique insights into early Hox evolution since cnidarians constitute an outgroup to the Bilateria 
It is currently a matter of debate whether cnidarians possess bona fide
Hox genes, and if so, whether the Hox code originated prior to the cnidarian-bilaterian divergence. In the last few years, several studies have suggested that cnidarians possess both anterior and posterior Hox genes, but they lack group 3 and central Hox genes 
(see for varying gene nomenclature). More recently, Kamm and co-workers have advocated two seemingly contradictory hypotheses: (1) that cnidarians possess anterior Hox genes, but instead of bona fide
posterior Hox genes, they possess a posterior Hox/Cdx like gene 
, and (2) that cnidarian genes related to bilaterian Hox genes “should be regarded as Hox-like but not as true Hox genes” 
. A more recent study by Chourrout and co-workers suggests that the cnidarian-bilaterian ancestor possessed two to three ParaHox genes as well as an Anterior and group 3-like Hox gene, each of which subsequently underwent independent radiations within the bilaterian and cnidarian lineages 
Nomenclature of Nematostella Hox and Hox-related genes.
With respect to the Hox code, a 2004 study on the expression of five candidate Hox genes in the anthozoan sea anemone Nematostella vectensis
found support for the existence of a Hox code in the cnidarian-bilaterian ancestor; multiple Hox genes appeared to be present, and they were found to be expressed in distinct territories along the primary body axis 
. In contrast, the aforementioned 2006 study by Kamm et al. 
concluded that the Hox code was a bilaterian invention based on what they regarded as the absence of central, group 3, and posterior Hox genes in Cnidaria, and on differences between the Hox expression patterns between the anthozoan Nematostella
and the colonial hydrozoan Eleutheria
In the current study, we employ novel analytical methods and present an extensive battery of new evidence from the sea anemone Nematostella vectensis
to address the origin and early evolution of Hox genes and the Hox code. This evidence includes phylogenetic analysis of eighteen distinct Hox-related loci from Nematostella 
, linkage analysis of these eighteen loci based on an assembly of the Nematostella
and extensive corroborating gene mapping studies; and developmental gene expression assays for 20 Nematostella
Hox-related genes, 12 of which have never been described before. For those genes whose expression has been previously described, we reveal previously unknown aspects of the spatiotemporal expression that are critical to interpreting Hox evolution.
Contrary to some recent reports 
, multiple lines of phylogenetic evidence support the hypothesis that both anterior and posterior Hox genes and two ParaHox genes were present in the cnidarian-bilaterian ancestor. Seven Nematostella
genes appear to be descended from the founding members of the Hox1, Hox2, and Hox9+ families. We provide more detailed transcriptional annotation of a genomic cluster comprising two ParaHox genes as well as one Hox1 family member, three Hox2 family members, an even-skipped ortholog, an HlxB9 ortholog, and a Rough ortholog 
During larval development, the putative Hox1, Hox2, and Hox9+ homologs are expressed in a number of distinct spatial domains that collectively account for practically the entire primary body axis, from the aboral to the oral extremity. Five of these candidate Hox genes (anthox7, anthox8, anthox8a, anthox6a, anthox1a) and one candidate ParaHox gene (NVHD065) are also differentially expressed along the secondary body axis, known as the directive axis. These genes are expressed in nested subsets along the directive axis, suggesting that Nematostella may be employing Hox genes to pattern both its primary and secondary axes. Phylogenetic mapping of gene expression patterns on a molecular phylogeny suggests that differential expression along the primary body axis is a primitive feature of the Nematostella Hox-related genes, while differential expression along the secondary body axis evolved afterwards.
Collectively, these data suggest that at least a rudimentary Hox code was operative in the cnidarian-bilaterian ancestor and that it played a role in patterning the animal's primary body axis (and possibly the secondary body axis as well). Moreover, strong stabilizing selection has been operating on this Hox code that has maintained certain core characteristics despite being deployed in a bewildering array of animal forms for over half a billion years.