1. AP4-Mediated Repression vs. Induction of Target Genes

Although AP4 was originally identified as a transcriptional activator for viral gene expression [2]^[1], many subsequent studies showed that it functions as a repressor of transcription for many genes. In the literature, there are several studies demonstrating an activating function of AP4 in the development of mammals. Badinga et al. showed that AP4 binds to the E-box motifs at the insulin-like growth factor binding protein-2 (IGFBP-2) promoter and acts as a trans-activator in luciferase reporter assay [5]^[2]. Another gene that is transactivated by AP4 is the dopamine β-hydroxylase gene (DBH), which is involved in the conversion of dopamine to norepinephrine in neural crest stem cells [6]^[3]. It was shown that the promoter of the DBH gene contains both AP4 and SP1 binding motifs within its promoter region. Although not binding to the promoter by itself, the transcription factor GATA-3 interacts with either AP4 or SP1 bound on the DBH promoter to facilitate gene activation. Interestingly, AP4 also binds to the promoter of the pancreatic amylase 2A gene, and DNase footprinting experiments showed that there are several AP1 binding sites near the AP4 binding motif. However, the authors did not investigate whether the binding of AP4 changes the expression of amylase 2A [7]^[4]. We also identified an enrichment of the AP1 and SP1 binding sites in the vicinity of promoters occupied by AP4 by using in silico methods and the AP4 ChIP-Seq results obtained in DLD-1 cells described above [3]^[5]. Whether the enrichment of AP1 and SP1 binding sites is linked to activation or repression by AP4 or whether it occurs in a cell-type specific manner is currently unknown.

In the aforementioned genome-wide analysis of genes regulated by AP4 in the colorectal cancer cell line DLD-1 [3]^[5], we found that the prevalence of AP4 acting as a repressor is higher than mediating gene activation. A total of 884 direct targets of AP4 were identified, and 530 of those were significantly down-regulated 24 h after ectopic AP4 expression, whereas 354 were induced by more than 1.5 fold. In addition, promoters repressed by AP4 showed an increased number of AP4 binding sites and were located closer to the transcriptional start site when compared to AP4-induced genes. As a result, AP4-repressed genes displayed stronger ChIP-Seq peaks at their promoter regions than activated genes. The differential regulation of target genes may be determined and mediated by numerous AP4-associated proteins, such as chromatin-histone deacetylases (HDACs), SWI/SNF-related, matrix associated, actin-dependent regulator of chromatin (SMARCs) proteins, and histone methyl-transferases (EHMTs) that were identified in a proteomic approach by Chen and our lab [8]^[6].

Notably, AP4 was shown to suppress gene transcription in many studies that focused on single AP4 target genes. For example, in human immunodeficiency virus (HIV-1) latently infected cells, AP4, along with histone deacetylase 1 (HDAC1), negatively regulates viral gene expression by binding to the HIV-1 long terminal repeats (LTR) within the viral promoter, and preventing access of the TATA-binding protein (TBP; TFIID) to the TATA-box [9]^[7]. Similarly, AP4 downregulates the transcription of the human papillomavirus type-16 E7 oncogene, which is required for the maintenance of the transformed phenotype, by binding to the P542 promoter [10]^[8]. In a screen for p53 activating factors, AP4 was identified as a repressor of the human homolog of the murine double minute 2 (HDM2) gene [11]^[9]. We could show, that the Glutamine and proline Q/P-rich acidic domain of AP4 is required to interact with numerous accessory proteins, such as HDAC1 and SP1, and thereby facilitates repression of the HDM2 gene [8]^[6].

Moreover, AP4 forms a protein complex with transcription corepressor geminin (Gem) to repress the temporal expression of the neuronal gene phytanoyl-CoA α-hydroxylase-associated protein 1 (PAHX-AP1) in the fetal brain. The AP4-Gem complex recruits silencing mediator of retinoic acid and thyroid hormone receptors (SMRT) and histone deacetylase 3 (HDAC3) to down-regulate the expression of neuron-specific genes in non-neuronal cells. During further development of the adult brain, expression of both AP4 and Gem declines, leading to higher expression of PAHX-AP1. Dual-specificity tyrosine-phosphorylation regulated kinase 1A (DYRK1A), which is involved in neurodevelopment, is repressed by the AP4-Gem complex, but, in the fetal brain of patients with Down syndrome, there is a decline in the expression of this complex, resulting in the transcription of the DYRK1A gene [12]^[10]. Furthermore, a decrease in the human angiotensinogen promoter transcriptional activity was observed when AP4 was overexpressed in an in vitro CAT assay [13]^[11]. This indicates that AP4 may repress the human angiotensinogen gene, which codes for the precursor of the vasoconstrictor hormone angiotensin-II in controlling blood pressure [14]^[12]. Many additional, genes repressed by AP4 were identified in the genome-wide study mentioned above (3): among them were CDH1, CD44, CLDN1, 4,7, GDF15, and OCLN.

The repression of AP4 target genes is achieved, at least in part, via the recruitment of chromatin-modifying enzymes. For example, AP4 recruits HDACs for transcriptional repression of HIV-1 [9]^[7] and HDM2 [8]^[6]. However, treatment with HDAC inhibitors is not sufficient to mitigate AP4-mediated repression of the HDM2 gene. Using the HDM2-P2 promoter DNA sequences containing AP4 binding site “CAGCTG” to pull down AP4 and its interacting proteins, we identified the chromatin-remodeling SWI/SNF complexes and histone H3K9 methyltransferases GLP/G9A (EHMT1 and EHMT2), in addition to the components of the HDAC complex [8]^[6]. These data imply that AP4 recruits multiple chromatin-modifying complexes for repression of its target genes. Since AP4 can act as both trans-activator and repressor, it remains to be determined whether the AP4 protein interactome changes in these two modes and what is the underlying mechanism that allows AP4 to switch between mediating induction versus repression at different target genes. One point to be mentioned is that the interpretation as to how AP4 controls normal developmental functions from chromatin-based assays should be cautious because immortalized and tumor cell lines may have altered chromatin states and the transcription factor occupancy may not reflect that of normal cells.

2. AP4 DNA Binding Sites as Targets for SNPs

Single-nucleotide polymorphisms (SNPs) are positions in the genome having single nucleotide variations that are present at a frequency of higher than 1% in the human population [15]^[13]. Many of the SNPs are linked to diseases. It is believed that single nucleotide changes in the genome may alter the binding of transcription factors and, hence, the expression of genes, which leads to the onset of diseases. Indeed, about one-third of the SNPs identified in genome-wide association studies (GWAS) overlap with a transcription factor binding site [16]^[14]. Previous studies showed that AP4 displays differential binding towards specific SNPs. For example, enrichment of AP4 is detected at the interleukin 2 receptor alpha (IL2RA) SNP rs12722522*C, which is associated with type 1 diabetes [17]^[15]. Another example is the SNP rs1800734 located at the promoter of the mismatch repair gene homolog 1 (MLH1) gene. Loss of MLH1 leads to impaired DNA mismatch repair (MMR), which causes microsatellite instability in some cancer types [18,19]^[16][17]. The binding of AP4 to the MLH1 promoter protects the region from DNA methylation at least in part via the enrichment of the zinc finger repressor protein CTCF [20]^[18]. Of note, a significant difference in the CTCF ChIP signal between the mutated and wild-type allele is not detected upstream of the rs1800734 allele at the position of a predicted CTCF binding site. Instead, CTCF is only enriched near the wild-type rs1800734 allele which is occupied by AP4. This observation correlates well with our previous finding that AP4 interacts with CTCF via its acidic region [8]^[6]. Interestingly, the rs1800734 G > A mutation disrupts the AP4 binding sequence, CAGCTG, located at the promoter of MLH1. The loss of AP4 binding leads to an increase in DNA methylation at the promoter region and the epigenetic silencing of MLH1 [21]^[19]. In another study, Liu and colleagues found that the rs1800734 G > A mutation in several colorectal cancer cell lines enhances the long-range interactions between rs1800734 and the promoter and 3′ UTR region of the doublecortin-like kinase 3 gene (DCLK3), resulting in an increase in the expression of this potential oncogene [22]^[20]. The authors reasoned that the rs1800734 G > A mutation disrupts the AP4 binding site and creates a binding motif for the E26 transformation-specific (ETS) transcription factor and thereby mediates activation of DCLK3.

3. AP4 Controls Cell Proliferation, Senescence, and Apoptosis

Mitogens interact with receptors presented on the cell surface and activate signaling pathways related to cell proliferation, survival and death [23]^[21]. Activation of the c-Myc gene represents a nodal point for most mitogenic signaling pathways. Activated c-Myc regulates numerous transcriptional programs that translate the incoming signals into cellular responses necessary for cell proliferation, such as protein and RNA synthesis, cell cycle activation and activation of glycolysis. Overstimulation of the mitogenic signaling pathways due to mutations can lead to the activation of cell cycle inhibitors, like p16 and p21, which may mediate cell cycle arrest or permanent arrest/senescence [24]^[22]. c-Myc, as well as other members in the Myc family of proteins, are known to be involved in the development of vertebrates [25]^[23], and when the expression is derailed, c-Myc causes cancer formation [26,27]^[24][25]. c-Myc is known to be a transcriptional regulator of AP4 [28]^[26], and some of the c-Myc functions are executed via the activity of AP4 (Figure 1). Similarly, in Drosophila, d-Myc regulates AP4 homolog Cropped (crp) to control the development of the terminal branching of Drosophila tracheal tubes and cell size. In addition, overexpression of crp results in increased cell size and nuclear size [29]^[27], which may presumably be due to endoreplication, a process similar to the cell cycle without cytokinesis.

Figure 1. AP4 as a mediator of the c-Myc function. AP4 is involved in many processes, such as the control of cell proliferation, which may be a part of developmental programs, such as CD8+ T cell clonal expansion and Drosophila tracheal terminal cell branching. AP4 can also regulate apoptosis, growth arrest, and senescence, in which p53 seems to be an important downstream player. Evidence from siRNA knockdown experiments shows that AP4 represses apoptosis.

Among all human AP4 controlled genes, the cell cycle regulatory genes and the apoptosis genes were enriched in a Gene Ontology (GO) analysis, suggesting that AP4 regulates cell proliferation and death [3]^[5]. It was shown that human c-Myc, a driver of cell proliferation, up-regulates AP4 expression to repress the transcription of the cyclin-dependent kinase (CDK) inhibitors p21 and p16, which inhibit the cell-cycle, suggesting that a critical level of AP4 is needed for correct cell cycle maintenance or prevention of its inhibition [28,30]^[26][28]. Indeed, mouse embryo fibroblasts (MEFs) derived from mice with a deletion of AP4 display premature senescence [30]^[28]. The effects of AP4 deletion were dependent on the up-regulation p16 and p21 since RNAi-mediated repression of these factors was able to reduce senescence. In addition, AP4-deficient MEFs ectopically expressing c-Myc and mutant RAS were resistant to transformation and did not form tumors in mice. Interestingly, ectopic expression of AP4 combined with mutant RAS resulted in tumors in p53-deficient MEFs but not in p53-proficient, wild-type MEFs. Since the combination of c-Myc and RAS expression was sufficient for tumor formation in wild-type MEFs, c-Myc presumably has additional oncogenic properties that are not mediated by AP4.

AP4 was also required for cell-cycle re-entry after mitogenic re-stimulation of starved MEFs with 10% serum, since AP4-deficient MEFs showed a strong delay in DNA-replication in this assay [31]^[29]. Interestingly, activation of AP4-ER, a tamoxifen-inducible fusion of AP4 to the estrogen receptor (ER), was only able to induce S-phase entry and proliferation in the presence of 1% serum but not at 0.25% serum in human diploid fibroblasts (HDF) [31]^[29]. This was mediated by the direct activation of Cyclin Dependent Kinase 2/CDK2 by AP4. However, activation of Myc-ER, is able to induce proliferation in low serum and requires AP4, since deletion of AP4 blocked Myc-induced S-phase entry [31]^[29]. In addition, in other studies, cell cycle arrest or decrease in cell number were observed after inactivation of AP4 using siRNAs [32,33,34]^[30][31][32]. Therefore, AP4 appears to be required, but is not sufficient, for induction of cell cycle entry and proliferation by mitogenic stimulation and after c-MYC activation.

Taken together, a certain level of AP4 expression is required for efficient cell cycle progression, but elevated AP4 expression is not sufficient to stimulate cell proliferation in mammals. A role of AP4 in cell cycle control and maintenance is also suggested by the observation that AP4-deficient MEFs showed a delay of 24 h in the onset of DNA replication in this assay and a defect in cytokinesis after mitogenic stimulation [31]^[29].

In addition to its role in cell cycle regulation and proliferation, AP4 also mediates apoptosis of mouse lymphomas treated with dexamethasone by binding to the promoter region and the activation of the cell death signal inducer (pro)caspase-9 to initiate apoptosis [35]^[33]. Tadakuma and colleagues showed that AP4 is expressed at a particular physiological level to maintain the basic level of pro-caspase 9 in the cells to prepare for any stress challenges. However, when the level of AP4 is lowered by knockdown of AP4 using siRNA [32,33]^[30][31] or knockout of AP4 in MEFs [31]^[29], cells also undergo apoptosis. It is not clear whether the diverging results are due to the use of different cell types or treatments of the cells. In contrast, activation of AP4 (in the form of AP4-ER after the addition of hydroxytamoxifen) was shown to be ineffective in inducing apoptosis [31]^[29]. Therefore, stress, such as transfection, is hypothetically needed together with a high level of AP4 to induce apoptosis, as we observed apoptosis only in cells transiently transfected with AP4 but not in the cells stably transfected with AP4 and induced with doxycycline (our unpublished data and the example of AP4-ER activation mentioned above). Further studies are required to validate this hypothesis.

In transformed (HeLa and HCT116 cells), as well as non-transformed cells (retinal pigment epithelial cells), it was observed that AP4 expression is reduced in the G2 phase due to the β-TrCP-mediated ubiquitination and degradation [36]^[34]. When the non-degradable AP4 (E135A/S139A) mutant was expressed in HCT116 cells, nuclear atypia and chromosome segregation were observed, and, subsequently, DNA damage response was activated, implying that an elevated, deregulated level of AP4 may lead to aberrant cell division and genomic instability.

Interestingly, our recent research shows that confluent retinal pigment epithelium (RPE) cells display senescence only when AP4 is over-expressed for more than three weeks in addition to a slowdown of cell proliferation [37]^[35]. RPE cells are immortal cells exhibiting cell-cell contact inhibition and proper tight and gap junctions [38]^[36], which make the cells suitable for studying how cells respond to a long-term tumorigenic challenge. AP4 is known to induce p53 expression directly by binding to the TP53 gene promoter to cause premature senescence [39]^[37]. In another study, we demonstrated that a consistently low level of ectopic c-Myc expression induces AP4-mediated senescence in post-confluent RPE cells [37]^[35]. However, an elevated and persistent expression level of c-Myc increases the number of apoptotic cells but decreases senescence. Using siRNA-mediated knockdown, AP4 was shown to be required for the c-Myc-induced senescence. At a higher c-Myc level, p53 is induced by both c-Myc and AP4 to initiate apoptosis rather than cellular senescence. This study suggests that the choice of whether a cell enters cell death or aging depends on the relative levels of c-Myc and AP4 [37]^[35].

4. AP4 in Adaptive Immunity

A sophisticated adaptive immune system is essential for antigen-specific, long-lasting protection against pathogens. The adaptive immune system consists of two parts, namely the cell-mediated and antibody-mediated immune responses. In the cell-mediated immune responses, activated antigen-specific T cells are generated to fight against foreign antigens, and correct temporal expression of CD4 and CD8 proteins is essential for cell-fate determination between CD4⁺ helper T-cells and CD8⁺ cytotoxic T cells [40,41]^[38][39]. Inside the thymus, the most immature T cells are CD4 and CD8 double negative (DN). After β-selection [41]^[39], CD4⁺ and CD8⁺ double positive (DP) cells are selected. DP T cells that express relevant T cell receptors are screened by positive selection and differentiate into CD4⁺ or CD8⁺ single positive (SP) cells [42,43]^[40][41]. Interestingly, AP4 was shown to affect the decision of T cells fate in part by mediating the down-regulation of CD4 synergistically with Runx proteins in T cells [44]^[42]. In DN T cells where CD4 expression is repressed, AP4 binds to the proximal enhancer of the CD4 gene. Consistent with the previous finding that AP4 interacts with HDAC1 [8]^[6], HDACs are thought to be one of the mediators in the AP4-mediated repression of CD4, as histone H3K9 hyperacetylation is detected at the proximal enhancer of the CD4 gene in AP4-deficient T cells [44]^[42].

Besides suppressing the expression of the CD4 gene, AP4 is also essential for maintaining the optimal proliferation and proper development of CD8 SP T cells. In activated CD8 SP T cells, c-Myc is essential for the initial activation of AP4 expression. Chou et al. found that AP4 is dispensable for the initial proliferation of activated CD8 SP T cells within the first 3 days after a viral infection, when c-Myc is expressed at a high level [45]^[43]. From day 4 to day 6 after infection, when c-Myc starts to decline and AP4 is still expressed at a high level in wild-type CD8 SP T cells, AP4 knockout cells have a reduced proliferation rate and cell size compared to wild-type cells. Subsequent microarray and ChIP-seq experiments showed that more than half of the c-Myc and AP4 ChIP-seq peaks were overlapped, and about one-fourth of the differentially expressed genes in AP4-knockout T cells are the shared targets of c-Myc and AP4. Gene ontology analysis showed that these shared targets are related to metabolism, transcription, and several translational pathways [46]^[44]. Importantly, the majority of the shared binding sites are still bound by AP4 at 5 days after viral infection when c-Myc expression declines. These data indicate that AP4 is required for the activation of genes to sustain the metabolic needs of CD8 SP T cells after c-Myc expression declines. Of note, the expression of a stabilized form of c-Myc (c-Myc T58A) rescues the decrease in proliferation and cell size in AP4 knockout T cells, but it only partially rescues the decrease in the number of terminally differentiated KLRG1⁺ CD8⁺ cells [46]^[44]. This implies that AP4 has its unique functions in CD8+ cell differentiation. In future studies, it would be interesting to focus on genes that are bound by AP4 alone but not c-Myc.

In the antibody-mediated immune response, antibody-secreting plasma B cells and memory B cells are generated in the germinal centers (GC) inside the lymphoid organs [47]^[45]. GC are dynamic microenvironments for the selection and maturation of B cells [48]^[46]. In the dark zone (DZ) of GC, B cells proliferate rapidly and undergo somatic hypermutation to generate random mutations in the variable domains of the B cell receptor (BCR). The cells then migrate to the light zone (LZ), where they interact with CD4⁺ follicular helper T cells. The B cell clones expressing high-affinity of antigen-specific BCR will be selected and migrate back to the DZ for further rounds of proliferation and somatic hypermutation [49]^[47]. As one of the mediators of the cell cycle, AP4 was shown to be essential for the proliferation and development of high affinity B cell clones in the DZ. Induced by c-Myc, which is expressed only in the LZ, AP4 is expressed in both LZ and DZ, and its expression in DZ is maintained by IL-21 [45]^[43]. B cell-specific AP4 deletion in mice results in a decrease in GC B cell number and GC size. B cells expressing AP4 in the dark zone also are more actively dividing and undergo more rounds of somatic mutation. In addition, B cell-specific AP4-knockout mice show a compromised immune response, as they are less capable of clearing up the viral load in serum after lymphocytic choriomeningitis virus-mediated chronic infection. Among the 520 upregulated genes in AP4⁺ DZ B cells, the majority of these genes are related to the metabolic and protein translational pathways. This implies that, similar to the roles of AP4 in CD8⁺ T cells, AP4 expression in DZ B cells is necessary for maintaining the metabolism of the cells. Another similarity between AP4 expressed in B cells and CD8⁺ T cells is that their expression is, in both, maintained by interleukins. In CD8⁺ T cells, the expression of AP4 protein drops after the withdrawal of IL-2 [46]^[44]. Similarly, IL-21 increases the expression of AP4 in cultured B cells in a dose-dependent manner [45]^[43]. It is not clear how interleukins regulate the expression of AP4, but the regulation is presumably mediated via post-transcriptional mechanisms, as the expression of AP4 encoded from the retrovirus in CD8⁺ T cells is also IL-2 dependent [46]^[44].