Selective antagonism of cJun for cancer therapy

Activator protein-1 (AP-1) designates a family of oncogenic transcription factors (TFs) that are integral components located at the end of a number of key signalling networks, controlling vital cellular processes such as differentiation, migration, proliferation and apoptosis [1‐6]. AP-1 functions as homo- or hetero-dimeric combinations of proteins in the Fos and Jun sub-families (a broader definition of AP-1 includes ATF and MAF sub-families) [7, 8]. As a dimer, AP-1 binds to cognate DNA sites within gene promotor elements to influence the expression of a range of target genes that include cyclin D1, FasL, SDF1, TNFα, proliferin and CD44 [9‐11]. This review will focus on cJun, an AP-1 family member which is found to be upregulated or overexpressed in a large number of cancers (for a list of cancers associated with specific AP-1 members, see Table 1) [47‐50]. cJun has since become a major focus for drug discovery, and its terminal activity within a number of pathways makes for a compelling target to ablate oncogenic signals that occur at any signalling level. This review will describe the role of AP-1 in general and how cJun specifically is dysregulated in various cancers. We describe antagonists from the literature, categorised according to the interaction surface within the cJun-DNA complex they target, as potential therapeutics against cJun dysregulation.

AP-1 proteins bind to DNA via their basic leucine-zipper (bZIP) domain (Fig 1a); comprised of an N-terminal DNA binding domain (DBD) and C-terminal leucine zipper (LZ). The LZ is the site of AP-1 dimerisation where an intermolecular interaction is facilitated by the formation of an α-helical coiled coil. This incorporates hydrophobic packing of i, i+7 repeating aliphatic hydrophobic residues (a position of the heptad repeat), and leucine residues (d position) which are flanked by polar and charged residues (e and g positions). The DBD forms an N-terminal extension of the α-helices from each protein to grip the DNA in a manner comparable to forceps. This inserts into the DNA major groove where basic sidechains interact favourably with the DNA phosphate groups, while key residues form specific hydrogen bonding contacts with bases within the recognition sequence [52, 53].

AP-1 proteins bind to a pseudo-palindromic DNA sequence (5’-TGA C/G TCA-3’) known as the 12-O-tetradecanoylphorbol-13-acetate (TPA) response element (TRE) [54‐56]. The DBD of each AP-1 protein within a given dimer binds to a separate half site (5’-TGA-3’) on the sense/antisense strands, separated by a C/G spacer base. The cFos-cJun heterodimer has shown binding promiscuity to single nucleotide variants (SNVs) of the TRE sequence with up to a 40-fold variation in affinity [53]. SNVs of the TRE site, present throughout the human genome, likely provide a layer of modulation to the ability of AP-1 dimers to alter the transcription of genes where they occur. AP-1 dimers may also bind to the cAMP response element (CRE, 5’-TGA CG TCA-3’) with a relevant though weaker affinity [57]. Crystal structures of the cJun homodimer interacting with both TRE and CRE site DNA are available (PDB codes: 1H7H for TRE, 1JNM for CRE). The binding affinities for these related DNA sites is controlled by AP-1 dimer composition, which in some cases leads to changes in binding site preference (particularly within the broader definition of AP-1 which includes Jun/Fos-ATF/MAF heterodimers) [58, 59]. In totality, this produces a regulatory system where the expression level of individual AP-1 proteins, and their subsequent nuclear transport, control the amounts of each AP-1 dimer available to bind [60‐62]. The sequence of a potential DNA binding site (TRE, CRE or SNVs thereof) will then define the affinity for each dimer present. This allows global alterations in AP-1 expression, and therefore dimer composition, induced by signalling events to produce a fine-tuned effect at a specific site of transcription.

All AP-1 family members have transactivation (TA) domains but only the Fos sub-family have been shown to have transrepression (TR) domains; found at their C-terminus (Fig 1b) [63‐67]. Generally, these domains have been identified by mutation/deletion of residues/regions though little further mechanistic study has been carried out. The transactivational activities of cJun, cFos and FosB are considered strong whereas activities for JunB, JunD, Fra-1 and Fra-2 are considered weak or repressive [50, 68]. This activity will be modulated by both the bZIP-DNA binding affinity and the activity of the domains themselves, which have some sequence diversity and therefore presumably differences in activity. As with DNA site specificity, the transactivation activity is modulated by AP-1 dimer composition.

The activity of AP-1 is controlled by the cellular context within which it is operating. This cellular context can be thought of in terms of AP-1 protein expression levels defining the prominent dimer compositions; the post translational modification of AP-1 proteins which can enhance or diminish activity; and the genetic organisation within a given cell as defined by epigenetics. The importance of cellular context is highlighted by the pro-apoptotic activity of cJun in neurons, in contrast with the anti-apoptotic activity of cJun in hepatocytes [68]. The various roles of AP-1 components have been the subject of significant study using gene knockout (KO) and transgenic mice, as discussed thoroughly in a review by Wolfram et al [50]. They describe the embryonic lethality of cJun, JunB and Fra-1 (showing they are indispensable) but KO of any AP-1 causes some detrimental effect such as osteopetrosis (cFos KO) or male sterility (JunD KO). AP-1 KO and transgenic mice were phenotypically diverse implying clear differences in the function of these proteins. In addition to observing the effect of AP-1 family protein KO, specific genes which are under AP-1 transcriptional control have been elucidated. AP-1 cellular control is mediated through activity on genes including cell-cycle regulators such as cyclin D1, cyclin A, apoptotic proteins such as FasL and TNF-α, chemokines such as SDF1 and many others [9‐11, 69, 70].

AP-1 proteins respond to numerous environmental and cellular stimuli including cytokines such as TNFα, hormones and neurotransmitters such as growth hormone, growth factors such as EGF, bacterial lipopolysaccharide (LPS), UV damage to DNA and reactive oxygen species [1‐5, 71‐78]. Each stimulus produces a signalling cascade which can alter the activity of AP-1 by changing transcription or by direct activation. The signal is passed through a mitogen-activated protein kinase (MAPK) cascade (sometimes initiated by a small G protein such as RAS or Rac) which terminally acts upon transcription factors. cJun N-terminal kinase (JNK) and p38 have both been shown to act in this manner by their phosphorylation of transcription factors including myocyte-specific enhancer factor 2C (MEF2C), activating transcription factor 2 (ATF2) and cJun [79, 80]. It is known that MEF2C, ATF2 and cJun itself are able to influence cJun expression [81‐84]. JNK increases cJun transactivation activity through phosphorylation of Ser63 and Ser73 within the TA domain [79]. JNK phosphorylation has also been shown to decrease ubiquitination-dependent proteasomal degradation of cJun which effectively increases cJun activity by virtue of increased cJun levels [85]. The array of interconnected components within the AP-1 signalling system allows the transduction of multiple pathways; balancing signals to define transcription and allow AP-1 to assert its role in cellular processes. Some cJun pathways that generalise aspects of AP-1 signalling are shown in Fig 2.

Although AP-1 dysfunction has been implicated in other pathological conditions including asthma and rheumatoid arthritis, cancer is the most prominently studied and is the focus here [47‐51, 86, 87]. Some hallmarks of cancerous cells which can be linked with AP-1 dysfunction are growth signal autonomy, angiogenesis, lack of apoptosis and uncontrolled cell migration. The functional role of AP-1 proteins has been shown to vary, resulting in a mixture of oncogenic and anti-oncogenic effects, depending upon cellular context and dimer composition [48]. For example, a cJun homodimer tends to have oncogenic activity however a cJun-JunB heterodimer can repress that function [88]. We may look to the specific case of breast cancer where cJun has been shown to be upregulated whereas FosB is downregulated [12, 13, 35]. In this context, therefore, AP-1 family members must be considered and treated individually. We focus on the role of cJun which has generally been shown to have an oncogenic effect. Early studies on the role of cJun in cancer showed sequence homology with known viral oncoprotein vJun, and overexpression of cJun was capable of malignantly transforming rat embryonic cells towards a cancer phenotype [67, 89, 90]. Although this alone did not imply cJun overexpression as an in vivo cause of human cancer, it was a strong indicator of the involvement of cJun and led to an intense period of further research.

Genetic studies can illuminate the target genes through which cJun exerts influence to produce hallmark cancer cell behaviours. cJun has been shown to regulate a range of genes involved in tumour development including cyclin D1 (upregulated to stimulate proliferation), Fas (downregulated to inhibit apoptosis), proliferin (upregulated to stimulate angiogenesis) and CD 44 (upregulated to stimulate invasiveness) [48, 91‐95]. However, the picture is not simplistic as cJun may also contribute anticancer effects, for example it upregulates BCL-2 interacting mediator of cell death (BIM) which stimulates apoptosis [96]. cJun appears to be best described as an oncoprotein but there is balance in its effects; tempered by cellular context.

The literature paints a broad picture of cJun dysregulation across a number of cancers and collectively shows that cJun is generally overactive and as such could be antagonised for therapeutic effect. Cancer related effects can be induced at any level of signalling, but as dysregulation is modulated at the transcriptional endpoint of cJun, antagonism there will ablate any oncogenic signalling in the pathway. Signalling pathways may also modulate multiple TFs and other processes, so attempts to therapeutically interact earlier in the pathway will increase the likelihood of nonspecific effects, which can be undesirable. We generally observe these issues when observing the range of natural products which have been investigated for chemoprotective or chemotherapeutic effects related to AP-1 dysregulation, such as resveratrol and harmal extract [121‐123]. They have the potential to produce beneficial outcomes but their mechanism is typically poorly defined; usually due to their action on multiple targets.

Although the focus here is cJun it is important to reiterate that some AP-1 components can produce oncogenic effects, and others are anti-oncogenic depending on cellular context and specific dimer composition. A therapeutic would seek to antagonise cJun oncogenic activity selectively while avoiding interaction with closely related AP-1 family members which are functioning normally or may be producing anti-oncogenic effects. Antagonism of cJun can ultimately be achieved by either preventing the bZIP domain from engaging with the DNA or by preventing the binding of the TA domain to any relevant interaction partners. Currently, the most attractive site for cJun antagonism is to directly modulate DNA binding at the bZIP (either by inhibition of dimerisation or by preventing the bZIP from docking to the DNA), since this has already received significant study, whereas little is known about binding at the TA domains of AP-1 proteins. To antagonise complex formation, it is important to study the binding surfaces which bring the functional structure together (Fig 3). This consists of an extended three component interface involving dimerisation of the LZ domains and the interaction of the DBD from each monomer with the DNA. In this section we shall describe a range of cJun antagonists (Table 2), categorised by the binding interface they target, to sketch a picture of the field.

AP-1 serves as a transcriptional super controller, transactivating target genes to modulate a variety of cell signalling pathways. These signals, from an array of sources, alter transcription to control processes such as differentiation, migration, proliferation and apoptosis. As such, oncogenic alteration of these pathways is also coordinated through AP-1 to produce cancer phenotypes through the promotion of growth signal autonomy, angiogenesis, lack of apoptosis and uncontrolled cell migration. A focus on cJun has highlighted the specific evidence of the role of this AP-1 family member in various cancers. The study of small molecules and peptides have both led to progress in the search for antagonists of this oncogenic cJun activity and provided important research tools to probe and further validate the role of cJun in cancer. There have been developments in a variety of surface targets, whether this is the TRE site of DNA or some stretch of the bZIP domain, though none have reached clinical use. This has particularly highlighted the importance of utilising the sequence diversity of the cJun LZ to specifically antagonise cJun oncogenic effect, to avoid antagonism of the potentially anti-oncogenic effect of other AP-1 family members. Functional antagonism, rather than non-functional cJun binding, may be best achieved by an antagonist which binds to the full bZIP domain to prevent both dimerisation and DNA binding simultaneously. With a range of challenges still to be overcome, any method to antagonise cJun-DNA binding may yet prove to be the route to a clinical therapeutic.