TFAP4 (AP-4)

The factors

TFAP4 (AP-4) is a single-gene family: there is one human TFAP4 protein and no close paralogue, which makes it one of the few bHLH families where the usual paralogue problem does not apply. It is a basic helix-loop-helix leucine zipper protein, but with an unusual architecture. Beyond the bHLH and the adjacent leucine zipper, AP-4 carries a second, more distal zipper-like dimerization motif; biochemical dissection showed that these multiple dimerization surfaces restrict it to homodimers and prevent it from partnering the other bHLH-ZIP proteins ^2123466. Functionally that puts AP-4 in the opposite position from the E-proteins: it does not form the combinatorial heterodimers that dominate the rest of the bHLH class, so its DNA-binding activity depends on its own expression level rather than on the availability of a partner. AP-4 was discovered as an enhancer-binding factor that cooperates with AP-1 to activate SV40 late transcription in vitro ^2833704, and it is expressed broadly but at levels that follow proliferative signals, because TFAP4 is a direct MYC target gene ^18818310. The atlas assigns 6 motifs here (4 JASPAR, 1 HOCOMOCO, 1 CIS-BP), all labelled TFAP4, with cohesion of only 0.77 - the lowest in this batch - for reasons explained below.

The motif

The AP-4 site is the palindromic class-A E-box CAGCTG, six base pairs, read as two CAG half-sites by the basic region of each monomer in the homodimer ^2123466. In the general bHLH scheme, CACGTG is the class-B E-box read by the MYC/MAX/USF/MITF group and CAGCTG (and CACCTG) is the class-A E-box read by the E-proteins and by tissue-restricted bHLH factors ^10611221; AP-4 is a class-A binder that happens to be a bHLH-ZIP. In the databases the motif is usually written extended, AaCAGCTGaT or CAGCTGtT, adding a weak preference for A/T immediately outside the core; that extra information is small and comes mainly from local DNA shape rather than from direct base contacts ^23562153. The low cohesion score has two concrete causes. First, two of the six models are not CAGCTG at all but CATATG-type CANNTG variants (aaCAtaTGtt in JASPAR MA1570.1, AhCATATG in CIS-BP), so the family contains two different E-box classes. Second, two of the models are JASPAR heterodimer profiles in which the TFAP4 half-site is fused to an ETS half-site - ACCGGAAACAGCTG with FLI1, cCGGAAGCAGsTG with ETV1 - which are composite models of a TFAP4-ETS pair rather than definitions of the AP-4 site itself. Across all four databases, the nearest neighbours of the TFAP4 models are usually motifs labelled with some other factor: label agreement is only 0.1-0.2. That is not an error in the databases; it is a statement that CAGCTG is not an AP-4-specific sequence.

Where it acts

CAGCTG is a 6-bp palindrome and therefore expected several hundred thousand times in the human genome, and unlike CACGTG it contains no CpG and so is not depleted. AP-4 occupancy is concentrated at promoters of cell-cycle and proliferation genes, overlapping heavily with MYC-responsive promoters, which is unsurprising given that TFAP4 is MYC-induced. It can act as an activator or a repressor at the same class of site; at CDKN1A it is a repressor ^18818310. The JASPAR composite models suggest AP-4 half-sites adjacent to ETS half-sites as one recurring arrangement, and the founding observation was cooperation with AP-1 at the SV40 late enhancer, so AP-4 sites are best interpreted as part of composite promoter modules rather than as stand-alone elements.

Biology and disease

AP-4 sits in a MYC-driven cell-cycle circuit. TFAP4 is induced directly by MYC and represses the CDK inhibitor CDKN1A (p21), so MYC can suppress p21 indirectly through AP-4 ^18818310; AP-4 is required for mitogen- and MYC-induced cell-cycle progression ^25261373, and it attenuates DREAM-complex-mediated repression downstream of p53 ^36831504. The relationship is not uniformly oncogenic: in B lymphocytes, MYC was found to suppress tumorigenesis through TFAP4-dependent restriction of stemness, so the direction of the effect is context-dependent ^34283887. There is no well-established Mendelian disease attributed to TFAP4 in the sources reviewed here.

Reading a hit

A CAGCTG match is close to uninformative about AP-4 on its own. The same core is the preferred site of TCF3, TCF4 and TCF12, of MYOD and the myogenic bHLHs, of ASCL and NEUROD proteins, of TWIST/HAND, and of many other class-A binders, and it is common enough to occur by chance roughly every few kilobases. Nothing in the flanking preference reliably separates AP-4 from those. Use a CAGCTG hit as "class-A E-box", then argue from expression, from ChIP data and from promoter context. Do not treat the two JASPAR TFAP4::ETS models as TFAP4 motifs; they are composite dimer models. Do not treat the CATATG models as the canonical AP-4 site either.