Downstream of the HOX genes: Explaining conflicting tumour suppressor and oncogenic functions in cancer

Abstract The HOX genes are a highly conserved group of transcription factors that have key roles in early development, but which are also highly expressed in most cancers. Many studies have found strong associative relationships between the expression of individual HOX genes in tumours and clinical parameters including survival. For the majority of HOX genes, high tumour expression levels seem to be associated with a worse outcome for patients, and in some cases, this has been shown to result from the activation of pro‐oncogenic genes and pathways. However, there are also many studies that indicate a tumour suppressor role for some HOX genes, sometimes with conclusions that contradict earlier work. In this review, we have attempted to clarify the role of HOX genes in cancer by focusing on their downstream targets as identified in studies that provide experimental evidence for their activation or repression. On this basis, the majority of HOX genes would appear to have a pro‐oncogenic function, with the notable exception of HOXD10, which acts exclusively as a tumour suppressor. HOX proteins regulate a wide range of target genes involved in metastasis, cell death, proliferation and angiogenesis, and activate key cell signalling pathways. Furthermore, for some functionally related targets, this regulation is achieved by a relatively small subgroup of HOX genes.

posterior embryonic axis in the same order, both spatially and temporally, as they are in the cluster, with the most 3 0 member expressed first and most anteriorly. This cluster has been duplicated twice in the course of evolution, such that mammals have four clusters, the members of which are also expressed in the same order along the anterior to posterior axis, with the 3 0 member expressed first, reflecting a closely conserved regulation of HOX genes between invertebrates and vertebrates. 3 In humans, the four clusters of HOX genes are named A, B, C and D, and are located on chromosomes 7, 17, 12 and 2, respectively.
The individual genes within each cluster are numbered consecutively from the 3 0 end, so, for example, the first gene in the HOXB complex is named HOXB1. The genes in the equivalent position in each of the other complexes are paralogous, having descended from the same ancestral HOX genes, thus HOXA4, HOXC4 and HOXD4 are paralogous of HOXB4. In general, proteins encoded by genes within the same paralogue group have similar sequences, and often have similar functions, 2,3 although a notable exception, HOXC10 and HOXD10, is described later in this review. A number of HOX genes have been lost during evolution, meaning that in total the four HOX clusters contain 39 genes. 3  context-dependent manner. In addition, HOX proteins have roles in the cytoplasm, although these are less well studied. A notable example is the binding of HOXA10 to p38 MAPK leading to its dephosphorylation and hence attenuation of p38 MAPK/STAT3 signalling. 5 Thus, the cellular location of HOX proteins is important in determining their function, and is in turn influenced by posttranslational modifications that regulate cofactor binding. 6 For example, acetylation of a lysine residue in HOXB9 (AcK27-HOXB9) causes it to become localised to the cytoplasm, preventing it from transcriptionally regulating its target genes. 7 In addition to their role in development, it is now well established that HOX genes are strongly expressed in most types of cancer, including prostate, 8 breast, 9 lung, 10 renal, 11 head and neck 12 and ovarian 13 cancer, as well as mesothelioma. 14 Although the expression of certain HOX genes is often reported as being characteristic of specific cancers, for example, HOXC4 and HOXC6 in prostate cancer, 8 generally the majority of the 39 HOX genes are significantly upregulated.
There have been numerous studies on the role of HOX genes in cancer, which can be divided broadly into associative studies (establishing a statistically significant relationship between HOX gene expression and key clinical parameters such as disease-free survival), and mechanistic studies, in which for example the downstream target genes of HOX proteins are identified with experimental evidence for a role in cancer development or progression, using molecular and cell biology approaches such as gene knock out, reduction of expression using short interfering RNAs, or gain of function experiments through the forced expression of a HOX gene. There are many more associate studies than mechanistic ones, although many mechanistic studies also include associative data.
To date, the overall picture that emerges from these studies is rather ambiguous-while some indicate that a particular HOX gene has a tumour suppressor function, others have ascribed it a prooncogenic function in the same cancer. In this review, we have focused on mechanistic studies in which molecular and cellular evidence is provided to support a tumour suppressor vs oncogenic function for different HOX genes. On this basis, the majority of HOX genes actually have a pro-oncogenic function. Our analysis reveals that the HOX genes activate (or more rarely repress) a wide range of genes involved in metastasis, cell death, proliferation and angiogenesis, and also activate key cell signalling pathways. While some of these functionally distinct groups of target genes are regulated by many members of the HOX family, others are activated by just a few paralogue groups. HOX genes reported to have both tumour suppressor and pro-oncogenic activities might be contextdependent with respect to cancer type and background gene expression.

| ONCOGENIC vs TUMOUR SUPPRESSOR FUNCTIONS OF HOX GENES
It is now clear from the growing body of work across multiple cancer types that while the majority of HOX genes seem to be associated with oncogenesis, a significant minority instead have tumour suppressor functions. This has made it more difficult to come to a global understanding of HOX function in cancer, as for example, some HOX genes are reported to have oncogenic functions in around half of the published studies, but tumour suppressor functions in other studies.
Excluding studies that are based only on associative studies (eg, the expression levels of a HOX gene compared to overall survival), and focusing instead on mechanistic studies where the function of the HOX gene in cancer is directly addressed, may help provide a clearer view (Table 1) As discussed above, HOXD10 is exceptional amongst the HOX genes in acting solely as a tumour suppressor-there have been no functional studies of its mechanism published to our knowledge that do not report tumour suppressor activity, with the exception of HNSCC in which it seems to reduce cell invasion but increase cell proliferation. 124 It is noteworthy though that its paralog in the C cluster,

| DIRECT HOX TARGETS
Amongst the publications that report mechanistic studies for the role of HOX genes in cancer, 28 provide experimental evidence for direct regulation of target genes (ie, HOX protein binding to the promoter/ enhancer regions that is required to activate or repress transcription).
Notably, all of these identify a pro-oncogenic role for the HOX protein(s) being studied ( Table 1). The directly regulated targets identified in these studies are summarised in Figure 5. With the exception of EGFR, which is upregulated by both HOXB5 160  PD-L2 can bind to the PD-1 receptor on T-cells to block their activation, 162 and TDO2 converts tryptophan to kynurenine, which also has an inhibitory effect on T-cells. 163 Elevated HOXC10 and TDO2 expression in glioma are both associated with immune suppression that supports tumour progression, indicating that HOX genes have functions in immunosuppression in addition to the other prooncogenic roles described above. 64

| IMPLICATIONS FOR CANCER TREATMENT
Individual HOX genes have been identified as possible therapeutic targets in cancer based on their pro-oncogenic function. However, there are a number of difficulties with this approach. One of these is the high degree of functional redundancy amongst HOX genes with respect to downstream target genes in cancer. Examples of these are discussed above but include genes involved in the EMT, where, with the possible exception of Claudin, target genes are regulated by two or more HOX proteins, and for E-cadherin by at least six HOX proteins. One approach to overcoming this is to target multiple HOX proteins through their interaction with PBX using an inhibitory peptide (HXR9) that acts as a competitive antagonist by mimicking the conserved hexapeptide region in HOX paralogue group proteins 1 to 10 that mediates PBX binding. 164 HXR9 has been shown to cause apoptosis in a range of cancers both in vitro and in vivo, including prostate, 8 breast, 9 lung, 10 renal, 11 ovarian, 165 oral 12 and oesophageal cancer, 166 as well as mesothelioma, 14 myeloma, 167 melanoma 168,169 and acute myeloid leukaemia. 170 A potential difficulty with this approach though is the apparent tumour suppressor function of some

| CONCLUSIONS
The role of HOX genes in cancer has now been extensively studied, although this has resulted in a number of apparently contradictory findings, especially the conflicting roles of HOX genes as tumour suppressors vs promoters of oncogenesis. In order to better understand the mechanisms by which HOX genes act in cancer, we have considered only those studies that provide experimental evidence for HOX function, rather than associative studies restricted to the relationship between HOX expression and clinical outcomes. On this basis, it seems that the role of HOX genes in cancer is predominantly prooncogenic, with the exception only of HOXD10. Furthermore, when grouped together by function, many target genes are regulated by only a few HOX proteins. This indicates that while HOX proteins are potential therapeutic targets in cancer, this may ultimately depend on an approach rooted in personalised medicine that adapts to the HOX expression profiles of individual tumours.

CONFLICT OF INTEREST
The authors have no conflicts of interest to declare.