Hox genes in development and beyond

Homeobox transcription factors are defined by a conserved 180 bp region of DNA, the homeobox sequence, that encodes a DNA-binding homeodomain. Phylogenetically, there are 11 groups of homeodomain-containing genes with Hox genes belonging to the ANTP group (Holland et al., 2007; reviewed by Ferrier, 2016). These genes were first described in the fruit fly, Drosophila (Lewis, 1978; Nüsslein-Volhard and Wieschaus, 1980), with subsequent studies identifying and characterizing them in a variety of other species, ranging from planaria to zebrafish to mouse and humans (reviewed by Feng et al., 2021; Ghosh and Sagerström, 2018; Mallo and Alonso, 2013; Olson, 2008; Quinonez and Innis, 2014). Distinguishing features of Hox genes in bilaterians are their linked clustering on the chromosome and their temporally and spatially directed expression (Duboule and Dollé, 1989; Izpisúa-Belmonte et al., 1991). Although we know that the temporal and spatial activation of Hox genes is essential for organismal development, many genomic targets of these transcription factors remain a mystery. In addition, how Hox protein structure, beyond the conserved homeodomain, and associated YWPM domain contributes to the control of gene expression is understudied. Here, we provide a brief overview of this important and interesting group of genes. We discuss the evolutionary history and the genomic organization of these genes and how this impacts cluster expression, Hox protein structure, and some key genetic functions described for Hox genes, with an emphasis on mammalian/vertebrate studies.

Homeobox-containing genes often encode transcription factors that play crucial roles in development and patterning. Significant work has been done to categorize the hundreds of homeobox genes into families and classes. The result of this work has established 11 homeobox-containing gene classes in animals (ANTP, PRD, LIM, POU, HNF, SINE, TALE, CUT, PROS, ZF, CERS) and 14 in plants (HD-ZIP I-IV, KNOX, BEL, PLINC, WOX, DDT, PHD, NDX, LD, PINTOX, SAWADEE) (Holland et al., 2007; Mukherjee et al., 2009). Hox genes are members of the ANTP class (Holland et al., 2007; reviewed by Holland, 2013).

Homeobox-containing genes are present in the three kingdoms of Eukarya (plants, animals and fungi), but Hox genes did not appear until after the divergence of the Porifera (sponges), evidenced by their presence in Cnidaria (corals, hydra, jellyfish, etc.) (Burke et al., 1995; Chourrout et al., 2006; Larroux et al., 2007; reviewed by Holland, 2013; Hrycaj and Wellik, 2016; Lemons and McGinnis, 2006). Cnidaria, however, lack a clustered arrangement of Hox genes; instead, Hox genes are randomly distributed throughout the genome, perhaps influencing the radial symmetry of Cnidaria (He et al., 2018; Kamm et al., 2006). Cnidaria and Bilateria likely diverged from a common ancestor, placing them on different evolutionary trajectories. Unlike Cnidaria, Bilaterian Hox genes are clustered on the chromosome and their expression patterns are under both spatial and temporal control. The order and developmental timing of expression of Hox genes along the chromosome mirrors their anterior-to-posterior expression, a property that has been referred to as colinearity (Duboule, 1994, 1998; Duboule and Dollé, 1989; Graham et al., 1989).

Although both invertebrates and vertebrates share colinear and spatiotemporal features of Hox gene expression, there are subtle differences in the genomic organization. In Drosophila, for example, eight Hox genes (lab, pb, Dfd, Scr, Antp, Ubx, abd-A and Abd-B) are linked, but the single Hox gene cluster is split across two different chromosomes (Von Allmen et al., 1996). In vertebrates, it is hypothesized that, in early vertebrate evolution, Hox genes were consolidated to form a compact, organized Hox cluster on the same strand of DNA, without non-Hox genes or repeat sequences. Global, regulatory control of the expression of these genes then evolved such that their chromosomal arrangement and structure became fixed (reviewed by Duboule, 2007). The most posterior member of the original cluster, Abd-B, expanded prior to duplication events, which resulted in additional Hox genes in vertebrates (39 in mammals and 47 in teleost), that are organized into four chromosomal clusters in mammals (HoxA, HoxB, HoxC and HoxD) and seven in zebrafish (Amores et al., 1998; Izpisúa-Belmonte et al., 1991; reviewed by Holland, 2013; Meyer and Málaga-Trillo, 1999; Singh and Krumlauf, 2022; Wagner et al., 2003). Each duplicated Hox gene in vertebrates is referred to as a paralog and this term refers to orthologous Hox genes within a species that have been classified based on genomic location, similarity of DNA sequence, and expression. For example, in mice, there are three members of the Hox1 paralogous group that reside on clusters A, B and D (Hoxa1, Hoxb1 and Hoxd1).

The initiation of Hox gene expression and the spatiotemporally controlled onset of expression relies on strict epigenetic regulation of Hox gene clusters. In vertebrates, from the early zygote stage until the initiation of gastrulation, Hox gene clusters are tightly packed and inaccessible, and are they are not expressed. Histone modifier complexes, such as the Trithorax (TrxG) and Polycomb (PcG) group complexes are important for regulating this chromatin environment (reviewed by Casaca et al., 2014; Gentile and Kmita, 2020; Jambhekar et al., 2020; Kassis et al., 2017; Mallo and Alonso, 2013; Noordermeer and Duboule, 2013; Schuettengruber et al., 2017; Schwartz and Pirrotta, 2007). As gastrulation commences, Hox expression progressively initiates, starting with the Hox1 and Hox2 paralogs in the newly forming anterior regions of the embryo (Forlani et al., 2003; Wacker et al., 2004). As gastrulation proceeds, the Hox gene complex becomes progressively available, promoting deployment of the next set of paralogs (Holland and Hogan, 1988; reviewed by Deschamps and Duboule, 2017; Deschamps and van Nes, 2005). Hox13 paralogs are turned on last, and this set of paralogs plays crucial roles in ending gastrulation along with sister homeobox genes from the Cdx family (Economides et al., 2003; Neijts et al., 2017; van de Ven et al., 2011; Young et al., 2009; reviewed by Mallo et al., 2010). Similarly, TrxG and PcG complexes regulate chromatin accessibility and segmentation in the developing fly (Bantignies et al., 2011; Bi et al., 2022; Cao et al., 2002; Loker et al., 2021; reviewed by Kassis et al., 2017).

Global enhancer sequences located outside the consolidated Hox gene clusters and the colinear arrangement of genes within the cluster control the pace of Hox gene deployment during axial segmentation (Crocker et al., 2015; Gould et al., 1997; Kmita et al., 2000; Miller and Posakony, 2020; Sharpe et al., 1998; Spitz et al., 2003; van der Hoeven et al., 1996; Zákány et al., 2001; reviewed by Afzal and Krumlauf, 2022; Mallo, 2018; Soshnikova, 2014). Genomes are organized into topologically associated domains (TADs); sequences within a given TAD will interact with higher frequency compared with those outside the TAD (Dixon et al., 2012). Both the HoxA and HoxD clusters contain TAD boundaries, which influence chromatin accessibility. During limb development, enhancers on either side of the TAD work to coordinate two transcriptional waves that permit limb patterning; the early wave patterns the stylopod and zeugopod, whereas the late wave patterns the digits (Montavon et al., 2011; Tarchini and Duboule, 2006; reviewed by Woltering and Duboule, 2010). The switch from early to late transcriptional waves for Hoxd13 is facilitated by enhancers posited in telomeric gene deserts in two TADs that reside outside the Hox gene clusters (Andrey et al., 2013). Single-cell assay for transposase-accessible chromatin (ATAC), chromatin immunoprecipitation sequencing (ChIP) and RNA-sequencing studies further supported the role of Hox13 in this transition, by demonstrating that Hoxa13 and Hoxd13 are required to permit an open chromatin conformation and, thus, the transition from early/proximal to late/distal limb patterning (Desanlis et al., 2020a,b; Sheth et al., 2016).

Although zebrafish and amniotes have similar TAD features, zebrafish TADs lack some enhancers, suggesting that these elements are important for the evolution of fins to limbs (Acemel et al., 2016; Freitas et al., 2012; Gehrke et al., 2015; Woltering et al., 2014; reviewed by Mallo, 2018). In Drosophila, recent work identified tethering elements, which function in conjunction with TADs to regulate developmental processes. Tethering elements facilitate the interactions between long-range enhancers and promoters, whereas TADs function to prevent unintentional interactions with enhancers (Batut et al., 2022). Enhancer sharing has also been documented in Drosophila, where one enhancer regulates the expression of two Hox genes (Eksi et al., 2018; Miller and Posakony, 2020).

Hox proteins share a 60 amino acid homeodomain that contains a helix-turn-helix DNA-binding domain, which binds to -TAAT- motifs (Desplan et al., 1985; Gehring et al., 1994; Passner et al., 1999). All Hox proteins bind this same consensus site with approximately equal affinity; thus, although the homeodomain is a defining and crucial feature of Hox transcriptional regulation, it does not impart transcriptional specificity (reviewed by Mann et al., 2009). Some Hox proteins additionally contain a hexapeptide (Hx) motif, comprising a highly conserved YPWM motif and variable linker region, immediately N-terminal to the homeodomain sequence. Hexapeptide domains are present in Hox paralogous groups 1-8, and paralogous groups 9 and 10 proteins have a similar motif that retains some structural and functional roles, but the hexapeptide motif is absent from paralogous groups 11-13 proteins. This motif plays crucial roles in TALE (three amino acid loop extension) co-factor selectivity to regulate a subset of Hox targets (Chang et al., 1995; Dard et al., 2019; LaRonde-LeBlanc and Wolberger, 2003; Merabet and Galliot, 2015; Merabet et al., 2003; reviewed by Casaca et al., 2014; In der Rieden et al., 2004; Merabet and Mann, 2016). Additionally, there are substantial differences in the size of N- and C-terminal regions of Hox proteins that are hypothesized to play roles in Hox specificity (Joshi et al., 2007; Liu et al., 2008; Papadopoulos et al., 2010).

Despite binding to the same DNA consensus site with equal affinity, Hox proteins have transcriptional specificity. One example of how Hox proteins achieve specificity is through interaction of the YPWM motif with TALE domain co-factors. Like Hox proteins, TALE proteins are highly conserved throughout evolution, and their interaction with Hox proteins increases Hox target selectivity (Bridoux et al., 2020; Dard et al., 2019; Slattery et al., 2011; reviewed by Kassis et al., 2017; Merabet and Mann, 2016; Moens and Selleri, 2006; Rezsohazy et al., 2015). There are two classes of TALE proteins: PBC proteins and MEIS proteins. Drosophila Extradenticle (Exd) and vertebrate Pbx (1-4) proteins make up the PBC class, whereas Drosophila Homothorax (Hth) and vertebrate Meis (1-3) and Prep (1 and 2; Pknox1 and 2) are MEIS class proteins (reviewed by Moens and Selleri, 2006). Although PBC/MEIS proteins also have non-Hox-related activities, they function together with Hox proteins to regulate the development of the hindbrain, limb, axial skeleton, hematopoietic system and other organs (Capellini et al., 2006; Dassé et al., 2012; Deflorian et al., 2004; Delgado et al., 2021; Di Rosa et al., 2007; Erickson et al., 2007; Khandelwal et al., 2017; Machon et al., 2015; Peifer and Wieschaus, 1990; reviewed by Moens and Selleri, 2006; Schulte and Frank, 2014). In zebrafish, for instance, pbx4 (Lazarus) is required for Hox1 and Hox2 paralog function in the control of motor neuron migration in the hindbrain (Cooper et al., 2003). During branchial arch development, Hoxa2 and Meis cooperatively bind DNA to specify the identity of the second branchial arch (Amin et al., 2015; Hunter and Prince, 2002). In Drosophila limb regulation, Exd and Hth enhance abdominal Hox protein Ubx, but not thoracic Hox protein Antp, to bind to the Dll leg selector regulatory sequence (Gebelein et al., 2002). In the mouse proximodistal limb axis, Meis1/2 serve as a co-factors for Hox proteins, responding to both FGF and retinoic acid gradients to pattern the developing limb and specify anterior-posterior identity, respectively (López-Delgado et al., 2021; Mercader et al., 2000). Accordingly, genetic elimination of Meis1/2 leads to limb defects with proximalization of the skeletal elements (Delgado et al., 2021, 2020).

In addition to the conserved homeobox region and the YPWM motifs, Hox proteins contain substantial protein sequences both N-terminal and C-terminal to these domains. These regions are of variable sizes, with the N-terminal region ranging from 96 amino acids (in Hoxa2) to 339 amino acids (in Hoxa10), and the C-terminal domain varying from eight amino acids (in Hoxb9) to 194 amino acids (in Hoxa3) in mice. In general, the N- and C-terminal regions are highly conserved among paralogs, but more divergent between paralogs. There is, however, remarkably little known about the function of these domains, although they have been shown to be important for transcriptional regulation in some systems (reviewed by Mann et al., 2009). For example, Six2 is a downstream target of both Hox11 and Hoxa2; loss of the Hox11 paralogous group genes results in loss of Six2 expression, whereas loss of Hoxa2 leads to increased Six2 expression (Kutejova et al., 2008; Wellik et al., 2002). These transcriptional activities rely on interactions with the N- and C-terminal domains of Hox11 and Hoxa2 (Yallowitz et al., 2009). In addition, it has been shown that the N-terminal region of Hoxa10 is necessary for rib suppression, and that mutations in the N-terminal regions of Hoxa13 and Hoxd13 result in limb and genital defects (Fantini et al., 2009; Guerreiro et al., 2012; reviewed by Brison et al., 2012). Recent work suggests that the C-terminal region and homeodomain also influence interactions with previously inaccessible chromatin (Bulajić et al., 2020). Given the lack of specificity provided by the homeodomain, the N- and C-terminal domains are likely responsible for much of the region- and tissue-specific transcriptional activities of Hox proteins. Interrogation of these domains and identification of additional binding partners with Hox proteins is understudied.

One of the key features of Hox transcription factors is their ability to impart different identities or morphologies along the axes in which they are expressed. During embryogenesis, Hox gene expression becomes spatially restricted along the anterior-posterior (AP) axis of the body (i.e. along the neural tube, axial skeleton and internal organs) and Hox9-13 genes additionally show spatial restriction along the proximodistal (PD) axis of the limbs (reviewed by Deschamps and Duboule, 2017; Mallo, 2018; Mallo et al., 2010; Wellik, 2009).

During gastrulation, the ectoderm forms and divides into the surface ectoderm (which gives rise to hair, nails and epidermis) and the neuroectoderm (which gives rise to the neural tube). As development proceeds, the neural tube becomes subdivided into forebrain, hindbrain, midbrain and spinal cord. Neural precursor cells in these developing areas differentiate to form nerve and glial cells specific to each functional region of the brain and spinal cord (reviewed by Jessell, 2000).

The developing hindbrain is divided into eight segments called rhombomeres, which give rise to cranial nerves. Hox1-4 paralogous group genes are expressed in the hindbrain and regulate rhombomere-specific development of hindbrain motor and cranial neurons; defects in paralogous groups 1-4 affect hindbrain segmentation and boundary formation (Barrow et al., 2000; Carpenter et al., 1993; Chisaka et al., 1992; Davenne et al., 1999; Gaufo et al., 2003; Gavalas et al., 1997, 2003; Mark et al., 1993; Prin et al., 2014; Rossel and Capecchi, 1999; Studer et al., 1998, 1996; reviewed by Ghosh and Sagerström, 2018; Krumlauf and Wilkinson, 2021; Parker et al., 2016; Parker and Krumlauf, 2020; Tümpel et al., 2009). Hox5-11 paralogous group genes are colinearly expressed in the spinal cord and play roles in defining motor neuron types derived from the phrenic, lateral, hypaxial, preganglionic and medial motor columns, whereas Hox13 is thought to act as a repressor of neuronal growth along the AP axis (Economides et al., 2003; Holstege et al., 2008; Hostikka et al., 2009; Jung et al., 2010; Lacombe et al., 2013; Lin and Carpenter, 2003; Philippidou et al., 2012; Tiret et al., 1998; van den Akker et al., 1999; Vermot et al., 2005; Wahba et al., 2001; Wu et al., 2008; reviewed by Joshi et al., 2021; Philippidou and Dasen, 2013).

In vertebrates, the mesodermally derived somites form alongside the neural tube during gastrulation and these important structures give rise to the dermis, skeletal muscles, tendons and axial vertebrae (reviewed by Williams et al., 2019). The vertebrate axial skeleton attains region-specific morphology as a result of Hox expression and function, subdividing regions from anterior to posterior into cervical, thoracic, lumbar, sacral and caudal domains. Within hours to days of the initiation of expression of each Hox paralogous group, the anterior boundary of expression for each Hox gene is fixed and vertebral axial fate and morphology are irreversibly established (reviewed by Mallo et al., 2010; Wellik, 2007).

Each Hox paralog group exhibits a high level of functional redundancy in patterning the axial vertebrae. It is only when the function of all alleles in a paralog group is lost that dramatic regional homeotic transformations occur (Horan et al., 1995a,b; McIntyre et al., 2007; van den Akker et al., 1999; Wellik and Capecchi, 2003). Loss of paralog function in the axial skeleton results in homeotic transformations. For example, mice with mutations in all three members of the Hox10 paralogous group do not develop normal lumbar vertebrae. Instead, the Hox10 paralogous mutants present with ribs that abnormally extend from the thoracic into lumbar and sacral regions (Wellik and Capecchi, 2003). Recent work demonstrated homeotic transformations in zebrafish for the first time when CRISPR-Cas9 was utilized to create loss-of-function alleles in each of the seven Hox clusters, highlighting a role for Hox genes in axial patterning in fish as well (Yamada et al., 2021).

Elegant support for the extent to which paralogs maintain redundant function has been demonstrated by the exact swapping of Hoxa3 and Hoxd3 coding sequences. Mouse lines were engineered such that the two exons of Hoxd3 exactly replace the exons of Hoxa3 (and the compliment replacement of Hoxd3 exons with those from Hoxa3). Hoxa3 and Hoxd3 single loss-of-function mutant mice displayed remarkably different phenotypes, yet the ‘swapped’ animals displayed no phenotypes, providing strong evidence that Hoxd3 and Hoxa3 are functionally equivalent when expressed appropriately (Chisaka and Capecchi, 1991; Condie and Capecchi, 1993; Greer et al., 2000). Similar results were subsequently obtained by swapping Hoxa1 and Hoxb1 coding sequences (Tvrdik and Capecchi, 2006).

Hox paralogs also redundantly participate in patterning many organs along the AP axis. Hox3 paralogous genes have been shown to be required for thymus, thyroid and parathyroid development (Condie and Capecchi, 1994; Manley and Capecchi, 1995). Hox5 genes play early roles in the growth and branching of the embryonic lung, and Hox6 genes are important in pancreas development (Hrycaj et al., 2015, 2018; Larsen et al., 2015). Hox10 and Hox11 genes are both important for proper development of the kidney, with loss of Hox11 leading to kidney agenesis (Wellik et al., 2002; Yallowitz et al., 2011). Hox9-13 paralogous groups impact the patterning of the reproductive tract in both males and females, and mutations in these genes result in dramatic transformations of the reproductive tract and lead to infertility (Cermik et al., 2001; Dollé et al., 1991; Ma et al., 1998; Raines et al., 2013; Taylor et al., 1999, 1997).

In humans, 11 syndrome-causing germline Hox mutations have been characterized. The known human Hox syndromes tend to present in the most distal areas (syndactyly of hands and feet, microtias, hypospadias and craniofacial abnormalities), likely because other Hox mutation-causing abnormalities are not easily visualized (i.e. internal abnormalities) (reviewed by Quinonez and Innis, 2014).

Limb buds emerge as mesodermal bulges adjacent from the lateral plate mesoderm. In vertebrates, the forelimb bud arises at the cervicothoracic transition and the hindlimbs develop at the lumbosacral transition, linking axial patterning to limb development (reviewed by Petit et al., 2017; Song et al., 2019). Three main limb skeletal domains form within the limb bud soon after its emergence. From proximal to distal, these are: the stylopod (humerus/femur), the zeugopod (radius and ulna/tibia and fibula) and the autopod (hand/forepaw and foot/hindpaw bones). Abd-B-related Hox genes, groups 9-13, pattern the morphology of each of these segments, with Hox9 and 10 paralogous group genes responsible for establishing stylopod morphology, Hox11 paralog group genes patterning the zeugopod, and Hox13 paralogs (specifically Hoxa13 and Hoxd13) patterning the autopod (Davis et al., 1995; Fromental-Ramain et al., 1996a,b; Wellik and Capecchi, 2003).

There is significant evidence accumulating in the literature of continued Hox gene expression in postnatal and adult animals. Hox expression in adults has been reported in the female reproductive tract, the dermis, lung, muscles, the hematopoietic system, synovial cartilage, nervous system, skeleton, and other tissues (Giampaolo et al., 1994; Hutlet et al., 2021; Kawagoe et al., 1999; Li et al., 2021; Pineault et al., 2002; Rux et al., 2021; Song et al., 2020; Taylor et al., 1997; Yoshioka et al., 2021; Yu et al., 2018). Moreover, the stromal cells within many organ systems show continued Hox expression from embryonic to adult stages (examples include the gut, lung, bone, pancreas, muscle and reproductive tract) (Garcia et al., 2020; Hrycaj et al., 2018; Larsen et al., 2015; Song et al., 2020; Taylor et al., 1997; Yahagi et al., 2004; Yoshioka et al., 2021; reviewed by Rux and Wellik, 2017). Intriguingly, Hox genes maintain their strict, regionally restricted expression that is established at embryonic stages.

In many cases, the function of Hox expression in adult tissues has not been explored, but in examples for which post-embryonic function has been assessed, work has shown that Hox genes continue to be important. The mouse skeletal system is one system in which adult Hox expression and function has been rigorously examined in recent years. The expression of Hox genes (visualized with a Hoxa11eGFP endogenous reporter, in one example) is absent from the cartilage condensations and osteoblasts that establish the radial and ulnar bones (Swinehart et al., 2013). Instead, Hoxa11eGFP expression is observed in the outer perichondral/periosteal layer that surrounds these bones and continues to be regionally expressed in the periosteal layer of the radius and ulna throughout life and is also visualized in the endosteum (inner bone lining) and in bone marrow stromal cells (Swinehart et al., 2013; Rux et al., 2016). This expression is maintained regionally in the zeugopod through postnatal and adult stages. The isolation of Hoxa11eGFP-expressing cells from adult bone revealed that these cells are progenitor-enriched mesenchymal stem/stromal cells, with progenitor activity in vitro (Rux et al., 2016).

The broader significance of this finding is supported by the demonstration that region-specific Hox expression in progenitor-enriched mesenchymal stem/stromal cells of other skeletal elements, including in human, continues throughout life (Bradaschia-Correa et al., 2020; Rux et al., 2016). Genetic lineage analyses in mice further supports this finding. Using Hoxa11-CreERT2 to induce lineage labeling at embryonic, postnatal or adult stages, Hoxa11-expressing cells were demonstrated to serve as a stable stem/progenitor population, contributing to chondrocytes, osteoblasts and bone marrow adipose throughout life (Pineault et al., 2019). Conditional loss of Hox11 function at adult stages resulted in progressive and severe defects in the remodeling of the radius and ulna during homeostasis, highlighting a continued role for Hox genes in the vertebrate skeleton (Song et al., 2020).

Extensive work supports the role for Hox genes in the development of the hindbrain and spinal cord (reviewed by Parker and Krumlauf, 2020; Philippidou and Dasen, 2013). However, additional recent studies document continued expression of Hox genes in neurons after birth (Coughlan et al., 2019; Karmakar et al., 2017; Lizen et al., 2017). Conditional loss of Hox2 genes in post-mitotic glutaminergic bushy cells of the anterior ventral nucleus permits proper tonography and sound transmission that is lost in null mutants, but results in defects in sound frequency discrimination, demonstrating a continuing role for Hox2 genes in shaping the neural circuitry of the mammalian auditory system during postnatal stages (Karmakar et al., 2017). Mis-regulation of factors in glutamatergic/GABAergic synapses and calcium signaling results from loss of Hoxa5 in postnatal stages (Lizen et al., 2017). As many Hox family genes have continued expression after birth in the spinal cord, it will be of great interest to explore their functional roles at postnatal and adult stages (Coughlan et al., 2019).

In the lung, Hox5 gene expression peaks 1-2 weeks after birth and conditional loss of Hox5 during postnatal development results in dramatic defects in alveologenesis, with expanded airways and bronchopulmonary dysplasia (Hrycaj et al., 2018). Although Hox5 expression decreases at adult stages, it remains present, and conditional loss of function in adult mice results in the rapid onset of an emphysema-like phenotype (Li et al., 2021). Potential post-embryonic roles for Hox genes in other organs and tissues are even less explored, but continued Hox gene expression supports this possibility. Substantial literature documents strong, posterior Hox gene expression (Hox9-13) in stromal cells of the adult female and male reproductive tracts and mutations in Hox9-13 lead to infertility (Ma et al., 1998; Mucenski et al., 2019; Oefelein et al., 1996; Taylor et al., 1998, 1997). Moreover, the mammary gland is developmentally regulated by the Hox9 paralog group, and expression of these genes continues in adult tissue (Chen and Capecchi, 1999). Pancreatic mesoderm strongly expresses Hox6 genes, and a subset of stromal cells maintain Hox6 gene expression throughout life, although the functional significance of this is unknown (Garcia et al., 2020; Larsen et al., 2015).

Although the hematopoietic system does not obey simple positional Hox gene expression, there are many publications documenting Hox expression and function in hematopoiesis and hematopoietic cancer; whether roles in this system are developmental or continuing is less clear (reviewed by Alsayegh et al., 2019; Bhatlekar et al., 2018; Feng et al., 2021). The association of Hoxa9 with acute myeloid leukemia is perhaps the most studied, but Hox mutations are associated with several other cancers, including breast, pancreatic, lung, liver and ovarian (Calvo et al., 2001; Kroon et al., 1998; reviewed by Feng et al., 2021; Morgan et al., 2017).

Post-metamorphic and adult expression of Hox genes have received less attention in organisms other than mammals, but recent work highlights emerging roles for Hox proteins in homeostatic processes. In Drosophila, changes in post-developmental Hox gene expression affect behavior, such as locomotion, the ability to self-right, and flight (Enriquez et al., 2015; Issa et al., 2022, 2019; reviewed by Buffry and McGregor, 2022; Joshi et al., 2021). Lin-39 (Hox4-5), in Caenorhabditis elegans, is required to maintain the terminal identity of cholinergic motor neurons (Feng et al., 2020). Finally, several Hox genes (hox1, lox5b, hox3a, hox3b and post2b) regulate different aspects of tissue segmentation and fission behavior during asexual reproduction in planaria (Arnold et al., 2021).

Despite crucial roles for Hox function in so many aspects of development, tissue homeostasis and regeneration/repair, there are fundamental gaps in our existing knowledge. How these transcription factors achieve specificity in transcriptional regulation represents important unanswered questions in the field. Our understanding of what proteins serve as co-regulators with Hox proteins in addition to the PBC-class proteins to potentially achieve this specificity remains limited. The regions N- and C-terminal to the homeodomain are sure to harbor important interacting domains and it will be essential to identify additional Hox partners. During development, what does positional information mean? What are the mechanisms by which Hox-regulated ‘region-specific’ information is translated in vivo?

Another underexplored area is how Hox genes continue to play important functional roles after embryogenesis. The recent identification of Hox-expressing cells as adult stem/progenitor cells in at least some organ systems emphasizes the importance of examining later functions in other tissues. With sequencing costs rapidly decreasing, obtaining ‘-omic’ data is becoming feasible and has the power to address at least some of these questions. The challenge of testing and providing evidence for new hypotheses that emerge will be more difficult – and more important. Elucidating the mechanisms of Hox gene regulation of the many morphogenetic processes they control will no doubt further our understanding of development and evolution and could provide important opportunities for novel regenerative therapies.