|Home | About | Journals | Submit | Contact Us | Français|
The Activating Transcription Factor 2 (ATF2) has been implicated in transcription and DNA damage control, through its phosphorylation by JNK/p38 or ATM/ATR, respectively. ATF2 activities have also been associated with skin tumor development and progression. Here we summarize our present understanding of ATF2 regulation, function and contribution to malignant and non malignant skin tumor development.
Activating transcription factor 2 (ATF2) is a member of the ATF/cAMP-response element-binding (CREB) protein family of basic region leucine zipper proteins (Maekawa et al., 1989, Nomura et al., 1993). ATF2 bind with high affinity to the octameric cyclic AMP-responsive element (CRE, cyclic AMP response element) TGACGTCA, similar to the TRE (TPA response element which is the classic site for c-Jun, TGAGTCA) (Hai et al., 1989). Under non stress growth conditions ATF2 exhibits a low level of transactivation because of an intramolecular inhibitory interaction in which the bZIP and activation domains are engaged (Li and Green, 1996). Inducers of ATF2 (kinases or association with viral proteins) disrupts this association and activates transcription (Fig 1). For its transcriptional program, ATF2 heterodimerizes with other members of the bZIP family including Fos, Fra2, c-Jun, CREB and ATF1, each of which causes a different degree of DNA bending (Kerppola and Curran, 1993). This generates diverse DNA structures which contribute to regulatory specificity among bZIP family proteins that can bind to the AP-1 site. To gain its transcriptional activities ATF2 requires phosphorylation or association with viral proteins, as first demonstrated for the adenoviral protein E1A (Flint and Jones, 1991, Hagmeyer et al., 1995, Nomura et al., 1993). Phosphorylation of ATF2 on threonines 69,71 is mediated by stress activated protein kinases JNK (Buschmann et al., 2000, Gupta et al., 1995, Livingstone et al., 1995) or p38/MAPK (Raingeaud et al., 1996). Thus, stimuli that activate these kinases, including exposure to pro-inflammatory cytokines, UV irradiation, DNA damage or change in ROS, are among the classic inducers of ATF2 transcriptional activity (van Dam et al., 1995). Further, ATF2 can be also activated by growth factors via the Ras-ERK pathway (which will phosphorylate Thr71) in concert with the RalGDS-Src-p38 kinases (that will phosphorylate on Thr 69) therefore allow cooperation between distinct signaling pathways in its transcriptional activation (Morton et al., 2004, Ouwens et al., 2002, Zhu et al., 2004). Viral protein products also activate ATF2 via inducing its phosphorylation by its upstream kinases, as demonstrated for Epstein Barr virus immediate early proteins BZLF1 and BRLF1 (Adamson et al., 2000), or via their own kinases as demonstrated for human vaccinia related kinase 1 (VRK1) which causes ATF2 phosphorylation on Thr 73 and Ser 62 (Sevilla et al., 2004). The role of diverse signaling in activation of ATF2 is also illustrated by the heterodimeric partners of ATF2, which are also activated in a stimulus-specific manner. Therefore, a particular stimulus can thus lead to the different ATF2 complexes and thereby activate distinct sub-sets of target genes (van Dam and Castellazzi, 2001).
Work in the fission yeast S. Pombe identified MAPK phosphatases as important players in the regulation of ATF2 transcriptional activity. Intriguingly, such phosphatases are subject to regulation by c-Jun, which is also activated by ATF2. Thus, ATF2 activation of c-Jun regulates expression of MAPK phosphatases affecting the p38 and ERK signaling pathways (Sprowles et al., 2005).
Transcriptional complexes containing ATF2 were shown to include the high mobility group protein HMG I (Y) which is required for assembly of ATF2-NF-κB complex. (Du et al., 1993). The scaffold protein Sin1 was shown to potentiate ATF2-p38 signaling by association with both components (Makino et al., 2006). These findings demonstrate the role of scaffold and regulatory proteins in ATF2 assembly of transcriptionally active complexes. Association between ATF2 and the retinoblastoma gene product was shown to play important role in the regulation of TGF-beta 2 gene (Kim et al., 1992), thereby implicating ATF2 in growth inhibitory signals. Consistent with these studies is the finding that ATF2-C-jun heterodimer orientation is important for assembly of functional beta interferon enhanceosome which also consists of interferon regulatory factor 3 (Falvo et al., 2000).
Other regulatory components that impact ATF2 transcriptional activities include histone deacetylase 3, a class I histone deacetylase which suppresses MAPK11-mediated ATF2 activation and represses TNF gene expression (Mahlknecht et al., 2004).
Regulation of ATF2 activity also depends on its subcellular localization and stability. Heterodimerization of ATF2 with c-Jun was shown to impact its nuclear localization (Liu et al., 2006), thereby pointing to the role of functional heterodimeric complex in subcellular localization. Modification of ATF2 associated protein is therefore expected to also alter subcellular localization. ATF7, (aka ATFa) is among the ATF2 associated proteins whose subcellular modification is dictated by sumoylation. Sumoylated ATF7 is primarily cytosolic resulting in inhibition of its transcriptional activities impairing its interaction with TAF12 in TFIID (Hamard et al., 2007). Via its heterodimerization with ATF2, sumoylated ATF7 may be also capable of affecting ATF2 localization. By analogy, c-Jun which was shown to impact ATF2 localization is subject to sumoylation which attenuates its transcriptional activities (Bossis et al., 2005). Thus, ATF2 activities may be affected due to altered subcellular localization, which may be affected by its heterodimerization partners.
An important layer in the regulation of ATF2 is its ubiquitination dependent degradation by proteasomes and association with UBC9 (Firestein and Feuerstein, 1998). ATF2 ubiquitination is facilitated by its association with non-active JNK (Fuchs et al., 1997), is dimerization dependent (Fuchs and Ronai, 1999) and is primarily targeted to the transcriptionally active form of ATF2, (Fuchs et al., 2000) thereby providing a mechanism to limit ATF2 transcriptional output. ATF2 stability and activity could be prolonged upon its association with c-Myc (Miethe et al., 2001). In Schizosaccharomyces pombe, a single member of the SAPK family called Sty1/Spc1/Phh1 stimulates gene expression via the Atf1 transcription factor, which is similar to the human ATF2 factor, and which binds to closely related DNA sequences. Sty1 phosphorylation of Atf1 is required for modulating its stability (Lawrence et al., 2007).
Analysis of ATF2 function is particularly complex due to the expression of distinct protein isoforms, either products of differential splicing or alternate promoter usage of the ATF2 gene. For example, several isoforms that share the DNA binding and dimerization domains at the C terminus of the protein but differ at their N termini have been described (Kara et al., 1990) (Georgopoulos et al., 1992, Kara et al., 1990). The alternative splicing mechanism may account for the partial deletion of the neo gene inserted in ATF2 to generate a knock out mice. As a result, this mouse is not a complete knock out and expresses an ATF-2 protein lacking the region between amino acids 277 and 326 (Maekawa et al., 1999).
The better characterized splicing form of ATF2 is ATF2-small (ATF2-sm) (Bailey et al., 2002) that contains the first and last two exons of the full-length protein (Bailey et al., 2002), and results in the expression of a 28 kDa protein which is ubiquitously expressed. Interestingly, despite that ATF2-sm lacks most exons coding for functional domains, it was found to be transcriptionally active. Moreover, a study on myometrial cells revealed that distinct sets of genes are transcriptionally regulated by ATF2 and ATF2-sm (Bailey and Europe-Finner, 2005).
ATF2 is ubiquitously expressed, with the highest level of expression being observed in the brain (Takeda et al., 1991). Hypomorphic Atf2 mutant mice, which still express an alternatively spliced ATF-2 protein, exhibit lower postnatal viability and growth, defects in endochondrial ossification, and a reduced number of cerebellar Purkinje cells (Reimold et al., 1996). Null Atf2 mutant mice die shortly after birth (Maekawa et al., 1999) and display symptoms of severe respiratory distress with lungs filled with meconium. These features are similar to those of a severe type of human meconium aspiration syndrome. There is increased expression of the hypoxia inducible genes suggesting that hypoxia occurs in the mutant embryos and that it may lead to strong gasping respiration with consequent aspiration of the amniotic fluid containing meconium. Impaired development of cytotrophoblast cells and a decreased level of expression of the platelet-derived growth factor receptor α, which is one of the ATF-2 target genes, are observed in the mutant placenta, suggesting a possible linkage of these events.
Homodimer ATF2 transcriptional activities are limited. Heterodimerization with other transcription factors is required for efficient transcriptional output. In general, it is possible to divide ATF2 target genes into (a) regulation of transcription factors and proteins engaged in stress and DNA damage response (b) regulation of genes associated with growth and tumorigenesis (c) regulation of genes important for maintenance and physiological homeostasis. Table 1 outlines known ATF2 targets within these categories. Among the transcription factors shown to be regulated by ATF2 are c-Jun and ATF3. ATF2 and ATF3 were shown to play important roles in protection of cells from ionizing irradiation (Kool et al., 2003). Along its role in the stress and DNA damage response, ATF2 has been implicated in the activation of a large set of genes important in drug resistance (Hayakawa et al., 2003). In addition, ATF2 impacts the expression of key processes that are required for stress response, including Nitric oxide synthase (iNOS) (Bhat et al., 2002). ATF2 has been also implicated in the regulation of the ER stress regulatory protein Grp78 (Chen et al., 1997), suggesting that its activation under certain conditions may also impact ERAD. Along ATF2’s role in stress and damage response is its effect on cell cycle regulatory proteins including cyclin A and cyclin D1 (Lewis et al., 2005, Nakamura et al., 1995, Shimizu et al., 1998).
In many cases, the c-Jun –ATF2 heterodimeric complex is sufficient to elicit efficient transcriptional activity. Constitutively active ATF2 mutant form confirmed activation of TNFα Fas ligand and c-Jun expression (Steinmuller and Thiel, 2003). However, inclusion of additional transcription factors in the ATF2 complex has been reported. For example, the ATF2/c-Jun and NFATp complex has been implicated in the regulation of tumor necrosis factor alpha gene transcription (Tsai et al., 1996). Data from JNK KO mice further corroborated the importance of JNK activation of ATF2 for transcriptional activation of TNFa (Ventura et al., 2003).
Atf1 (also known as Gad7 or Mts1) is the S.Pombe homologue of the mammalian ATF-2. Atf1 is a b-ZIP protein that associates with- and is phosphorylated by- Sty1 following different stresses. In S.Pombe the central element of the SAPK cascade is the MAPK Sty1 (also known as Spc1 or Phh1), which is highly homologous to mammalian p38 kinase and becomes activated by different stresses. Atf1 protein forms heterodimers with Pcr1 (Mts2) protein (Wahls and Smith, 1994, Kanoh et al., 1996), but each of these proteins can also form homodimers that bind to DNA (Wahls and Smith, 1994). Atf1 and Pcr1 harbor basic leucine zipper (bZIP) motifs characteristic of dimeric DNA-binding transcription factors of the CREB/ATF family. Atf1 transcription factor is required for a variety of stress responses, including those caused by nutritional starvation, oxidative stress, osmotic stress, cold stress, UV damage, and nucleotide pool depletion ((Watanabe and Yamamoto, 1996, Kanoh et al., 1996, Degols and Russell, 1997, Hirota et al., 2004, Madrid et al., 2004, Quinn et al., 2002, Shiozaki and Russell, 1996, Soto et al., 2002, Takeda et al., 1995, Wilkinson et al., 1996).
ATF2 has also been implicated in the regulation of other signaling pathways, including TGFβ. Activation of matrix metalloproteinase (MMP)-2 by TGFβ was shown to be dependent on p38 and ATF2 (Kim et al., 2007, Song et al., 2006). Along these lines, TGFβ activation of Smad7 was also shown to be dependent on ATF2 (Uchida et al., 2001). Studies from ATF2 mutant mice revealed an impaired induction of the adhesion molecules E-selectin, P-selectin, and VCAM-1, as well as the cytokines tumor necrosis factor-alpha, IL-1beta and IL-6 following LPS treatment (Reimold et al., 2001) pointing out the importance of ATF2 for induction of adhesion molecules and cytokine genes.
ATF2 was also shown to regulate transcription of the tyrosine hydroxylase gene, a rate limiting factor in catecholamine biosynthesis, which is important during neuronal development (Suzuki et al., 2002). Collagen gene expression in osteoblasts has been shown to be regulated by ATF2 (Matsuo et al., 2006). ATF2 has been also implicated in the regulation of insulin promoter (Hay et al., 2007). Recently ATF2 and histone deacetylase 5/myocyte enhancer factor 2 has been implicated in the regulation of the Pgc-1alpha gene in skeletal muscle (Yan et al., 2007). Table 1 summarizes some of ATF2 transcriptional targets.
Using DNA microarrays of promoter sequences ATF2 was found to affect transcription of around 180 genes following treatment with cisplatin (Hayakawa et al., 2004). In most of the cases, c-Jun was found to be the ATF2 partner and changes in transcription were in the positive direction (Hayakawa et al., 2004). These studies points to the important and well defined role of ATF2 in transcription.
In addition to its function as a transcription factor, ATF2 was found to play an important role in DNA damage response. This recently identified function requires ATF2 phosphorylation by PIKK (ATM, ATR or PI3K kinases), on two serine residues within the extreme C-terminal domain (Serine 490,498 of hATF2). ATM phosphorylation of ATF2 was shown to cause its localization into DNA repair foci, where it co-localizes with components of the DNA repair machinery, Rad50, Nbs1 and Mre11 (Bhoumik et al., 2005). ATF2 phosphorylation by ATM, but not its phosphorylation by JNK or p38, is required for intra-S-phase checkpoint following ionizing irradiation. Thus, independent of its function as a transcription factor, ATF2 serves as an important component of the DNA damage response (Fig 2). Interestingly, while ATM is required for the phosphorylation of ATF2 and its function following DNA damage, ATF2 also contributes to the degree and level of ATM activities. This reciprocal relationship was found to be mediated by TIP60- a histone acetyltransferase. ATF2 was previously shown to associate with components of the TIP60 complex, including TIP49b (Cho et al., 2001). In more recent studies, other components of the TIP60 complex, including TIP60, were found to associate with ATF2. Interestingly, under non stress conditions ATF2 limits TIP60 stability via recruitment of the Cullin3 ubiquitin ligase to the complex. Following IR, ATF2 dissociates from TIP60, thereby gaining its expression and enabling increased ATM activity (Bhoumik et al., unpublished observations) by TIP60-dependent acetylation of ATM (Sun et al., 2005). Changes in ATF2 expression, as often seen in human cancer, coincide with altered expression of TIP60 and degree of ATM activation (Bhoumik et al. unpublished studies). Intriguingly, earlier studies in S. Pombe implicated Atf1/Pcr1 (ATF2 homolog) in meiotic recombination hot spots (Steiner et al., 2002). Furthermore, atf1 and pcr1 are required for deacetylation of histone H3 (and H4), a prerequisite for subsequent H3 lysine 9 methylation and Swi6-dependent heterochromatic assembly (Jia et al., 2004, Kim et al., 2004). These observations suggest that via its effects on TIP60 availability and concomitant activity, ATF2 affect chromatin organization and ATM activity. ATF2 was also shown to affect transcription of large set of genes implicated in DNA repair (Hayakawa et al., 2004) suggesting that its role in the DNA damage response require its dual functions as transcription factor and DNA damage response protein. It is yet to be determined whether any of ATF2 functions dominate in cells that are subjected to DNA damage when kinases that activate its transcription (JNK, p38) and DNA damage response (ATM, ATR) are co-activated. Among the possible considerations for function related activity is the unexplored fact that ATF2 is subject to extensive splicing processing. Given the distinct domains required for each of ATF2 functions, some of these spliced forms are expected to function as DNA repair but not transcription, others in transcription but not repair.
Studies from the laboratory of Ishii in Japan have provided two independent observations which point to the possible importance of ATF2 in hypoxia. First, this group has shown that mice in which ATF2 has been deleted exhibit increased expression of hypoxia inducible genes including VEGF (Maekawa et al., 1999). Along these lines, gene profiling array performed on hypoxic samples identified a large subset of genes whose expression was altered due to the absence of ATF2 (Maekawa et al., 2007). This suggests an important role for ATF2 in regulation of gene expression under hypoxia. Interestingly, hypoxia was also reported to induce expression of mitogen-activated protein kinase phosphatase-1 which was shown, via its effect on JNK, to impact c-Jun and ATF2 transcriptional activities (Laderoute et al., 1999). The emerging picture is that changes in regulatory components of the ATF2 signaling pathway take place under hypoxia to affect ATF2 activities and consequently the activities of the proteins which are subject to its transcriptional regulation.
The nature of ATF2 transcriptional targets which play a central role in cell cycle control, drug resistance and apoptosis, combined with changes reported for ATF2 kinases in tumor development and progression, raise the possibility that ATF2 may play important role in tumorigenesis. ATF2 cooperates with v-Jun to promote growth factor-independent proliferation in vitro and tumor formation in vivo (Huguier et al., 1998). Our own studies have pointed out the important role ATF2 plays in melanoma development and resistance to treatment (Ronai et al., 1998). In part, ATF2 mediates its effect in melanoma cells by altered p38 signaling which impacts Fas and NF-κB expression (Ivanov and Ronai, 2000). ATF2 also affects the expression of tumor necrosis factor alpha which acquires melanoma resistance to cell death by eliciting pro-survival signaling (Ivanov and Ronai, 1999). Consistent with the role of ATF2 in melanoma is the finding made in TMA of melanoma tumors. Analysis of >500 melanoma tumors revealed that nuclear localization of ATF2 associates with poor prognosis (Berger et al., 2003). Nuclear ATF2 is likely to be transcriptionally active whereas cytoplasmic ATF2 represents the inactive form. In an approach to selectively inhibit ATF2 activity in human melanoma we have designed peptides that derived from ATF2 transactivating domain, and tested their ability to affect ATF2 transcriptional activity. A peptide corresponding to aa 51–100 of ATF2 was able to inhibit, whereas a peptide corresponding to aa 151–200 increased ATF2 activities (Bhoumik et al., 2001). The latter are consistent with the notion that aa 151–200 are part of the domain required for ATF2 intra-molecular inhibition, thereby constitutive expression of this peptide interferes with this inhibition and renders ATF2 active. In contrast, ATF2 activation by JNK and p38 are mediated through sequences positioned within the 51–100 region, therefore such peptide is expected to outcompete activation of endogenous ATF2 to attenuate ATF2-dependent transcription. Of interest, expression of such the 51–100 peptide was able to sensitize melanoma cells to apoptosis following treatment with chemotherapeutic drugs (Bhoumik et al., 2001). Further, constitutive or inducible expression of the ATF2 51–100aa peptide in mouse melanoma SW1 cells efficiently attenuated tumor growth in syngeneic mouse model (Bhoumik et al., 2002). Adenoviral delivery of ATF2 peptide was also efficient in inhibition of mouse melanoma B16F10 growth using the corresponding mouse strain (Bhoumik et al., 2002). Delivery of HIV-TAT fused in frame with ATF2 peptide also elicited efficient inhibition of melanoma tumor growth (Bhoumik et al., 2004a). Efforts to minimize the size of this peptide succeeded, and 10 aa peptide which encompasses the JNK association site on ATF2 was found sufficient, albeit not as efficient as the longer 50aa in sensitizing melanoma cells to apoptosis and inhibiting their growth in mouse models (Bhoumik et al., 2004a). Mechanistically, the ATF2 peptides elicit their activity by two major means. First, they inhibit ATF2 transcriptional activities. Second, an important finding was that they also associate with JNK, and increase basal JNK activity. Thus, in the presence of the ATF2 peptide, JNK and its substrate c-Jun are activated, while ATF2 transcriptional activities are attenuated (Bhoumik et al., 2004b). In the absence of ATF2, JunD was found to cooperate with c-Jun and increase AP1 activities (Bhoumik et al., 2004b). Consistent with these findings, mutation of aa critical for JNK association on ATF2 peptide impaired its ability to sensitize melanoma to apoptosis and inhibit their growth in mouse models (Bhoumik et al., 2004b). These observations establish the ability to sensitize and consequently inhibit melanoma growth using ATF2-driven peptides. In efforts to identify compounds that may mimic ATF2 peptide activity we have screened a library of >3000 natural compounds. Two compounds that are capable of inducing apoptosis while inhibiting ATF2, activating JNK and c-Jun were identified. These compounds are also capable of eliciting efficient inhibition of melanoma growth and metastasis formation in mouse models (Abbas et al., 2007, unpublished observations). Consistent with the role of ATF2 in melanoma is the finding that the HGF transgenic mouse model which can be induced for melanoma development upon exposure to UV-B irradiation mediates melanoma proliferation via ATF2 (Recio and Merlino, 2002). Important support for these observations emerges from the mouse model in which ATF2 has been deleted in melanocytes, and upon cross with mice that spontaneously develop metastatic melanoma (N-RasTg/Ink4aΔ) were no longer able to develop melanomas (our unpublished observations). These studies establish the important role ATF2 plays in the development and progression of melanoma and points to its oncogenic function in this tumor type.
The possible activities of ATF2 as a tumor suppressor emerge from two independent studies. In the first, Ishii’s lab in Japan has demonstrated that ATF2 heterozygous mice were prone to develop mammary tumors, in particular upon their cross with p53 mutant mice (Maekawa et al., 2007). In part the tumor suppressor phenotype was associated with reduced level of Gadd45a and maspin expression, which are associated with apoptosis. The tumor suppressor function was also associated with altered expression of hypoxia inducible genes (Maekawa et al., 2007).
Work performed in our lab has identified tumor suppressor activities of ATF2 in keratinocytes and skin tumors. Inactivation of ATF2 transcriptional activities in keratinocytes, (K14-Cre) was found to increase incidence of skin papillomas in the 2 stage skin carcinogenesis protocol (DMBA-TPA). Lack of transcriptionally active ATF2 also shortened the latency for papilloma development. Molecular analysis revealed that in the absence of ATF2 there has been decrease in the expression of presinillin1 with concomitant decrease in Notch1 expression and increase in β-catenin levels (Bhoumik et al. unpublished observations). Both β-catenin and Notch were implicated in the development of skin papillomas and in skin tumors (Gustafsson et al., 2005, Kaidi et al., 2007). Thus, through its regulation of presinillin1, ATF2 was found to regulate an important upstream regulatory component required for skin cancer development.
These finding raises the important question regarding ATF2 ability to serve in certain tissues as an oncogene, and in others as a tumor suppressor. Examples for important transcription factors that elicit such opposing functions exist, and include Notch and β-catenin, although the mechanism underlying their ability to elicit diverse function is not completely clear. Among the possibilities to consider is the possible change in ATF2 transcriptional activities in each of these tissue types. Such change could be attributed to the tissue specific subset of heterodimeric partners, changes in ATF2 kinases, and possibly ATF2 subcellular localization. Analysis of TMA from human melanoma patients identified that nuclear localization of ATF2 coincides with poor prognosis (Berger et al., 2003), implying a role for constitutively active ATF2 in the more aggressive melanomas. Conversely, analysis of TMA from human skin cancer cases, including basal and squamous cell carcinomas identified a marked decrease in nuclear localization of ATF2, compared with the normal skin samples, which coincided with increase localization of ATF2 in the cytoplasm (Bhoumik et al. unpublished studies). Thus, principle differences in the expression of ATF2 among the tumor types may point to the mechanism underlying the opposing functions of ATF2 in benign and malignant skin tumors.
Present knowledge positions ATF2 as important transcription factor and DNA damage response protein, which is also implicated in the regulation of cellular growth control. ATF2 elicits its functions in intra-S phase checkpoint control in addition to its role in transcriptional regulation via distinct upstream regulatory kinases. Whether different pools of ATF2 can be utilized for each of these functions, or whether performing along the DNA repair axis impacts the transcriptional activities of ATF2 are issues that remain to be clarified. Along the growing complexity of understanding ATF2 regulation and function are the observations that point to its ability to elicit oncogenes or tumor suppressor functions, depending on the tissue type. Future studies will reveal the nature of these principal differences, and further delineate the important role ATF2 plays in cellular growth control prior and following DNA damage, as in transformation and cancer development.
Studies described in this review were performed with NCI grants CA099961, CA051995 (to ZR).