Upstream stimulating factors: highly versatile stress-responsive transcription factors

Authors


*Address correspondence to Marie-Dominique Galibert,
e-mail: mgaliber@univ-rennes1.fr

Summary

Upstream stimulating factors (USF), USF-1 and USF-2, are members of the eucaryotic evolutionary conserved basic-Helix-Loop-Helix-Leucine Zipper transcription factor family. They interact with high affinity to cognate E-box regulatory elements (CANNTG), which are largely represented across the whole genome in eucaryotes. The ubiquitously expressed USF-transcription factors participate in distinct transcriptional processes, mediating recruitment of chromatin remodelling enzymes and interacting with co-activators and members of the transcription pre-initiation complex. Results obtained from both cell lines and knock-out mice indicates that USF factors are key regulators of a wide number of gene regulation networks, including the stress and immune responses, cell cycle and proliferation, lipid and glucid metabolism, and in melanocytes USF-1 has been implicated as a key UV-activated regulator of genes associated with pigmentation. This review will focus on general characteristics of the USF-transcription factors and their place in some regulatory networks.

USF genes

The upstream stimulatory transcription factor (USF) family was first identified in human cells because of their involvement in the transcription regulation of the adenovirus major late promoter (Carthew et al., 1985; Moncollin et al., 1986; Sawadogo and Roeder, 1985). The complete purification of USF, from HeLa cell nuclei gave two related polypeptides of 43 and 44 kDa (Sawadogo, 1988; Sawadogo et al., 1988). Microsequencing analysis of the 43 and 44 kDa proteins resulted in the identification of human USF-1 and USF-2 genes respectively (Gregor et al., 1990; Sirito et al., 1992, 1994). USF-1 orthologues have been identified in several species, including yeast (Atchley and Fitch, 1997; Ferreiro et al., 2004), Xenopus (Kaulen et al., 1991), sea urchin (Kozlowski et al., 1991), mice (Henrion et al., 1996) and rat (Galibert et al., 1997). Sequence analysis has revealed that USF-1 is evolutionarily well conserved with Xenopus and sea urchin USF-1 cDNA showing 88% identity while the human USF-1 cDNA, and mice, rat and human USF-1 cDNA show up to 90% identity. The similarity is even higher in the proteins, with about 99% identity between rat, mice and human USF-1 proteins (Galibert et al., 1997). Comparable results were observed for orthologous USF-2 genes (Henrion et al., 1995; Lin et al., 1994), with murine USF-1 and USF-2 cDNA having 59% identity (Figure 1A) (Lin et al., 1994; Sirito et al., 1994).

Figure 1.

Gene and protein structure of the ubiquitous transcription factors, upstream stimulating factor (USF). (A) USF-1 and USF-2 genes constitute 10 exons. Sequences in colour correspond to the basic domain (yellow), the Helix-Loop-Helix (red) and Leucine-Zipper (black) (b-HLH-LZ) domains. The exon 4-excision form of USF protein members correspond to dominant negative regulators of the USF-transcription factors. (B,C) USF-1 and USF-2 belong to the b-HLH-LZ transcription factors and display similar protein structure. USR domain (USF-specific region), located upstream from the basic region is essential for transcriptional activation of promoters containing both the TATA box and the initiator element. USF-1 phosphorylation on T153 is necessary for transcription activity.

In mammals, USF-1 and USF-2 have a similar genomic organization, each comprising 10 exons (Figure 1A) spanning 4 kb for USF-1 and 10 kb for USF-2 (Henrion et al., 1995, 1996; Lin et al., 1994) with S1 nuclease mapping determining the transcription start sites. Both promoters lack an obvious TATA box motif, but an initiator element (Inr) consistent with the Inr pyrimidine consensus sequence (YYAYTCYYY) (Breathnach and Chambon, 1981; Smale and Baltimore, 1989) could be found. In addition, the USF-2 gene promoter, but not the USF-1 promoter, is characterized by two E-box motifs located at −332 and −186 from the transcription start site, potentially allowing regulation of the USF-2 gene by USF factors (Lin et al., 1994; Sirito et al., 1998).

Shieh et al. (1993) located the human USF-1 gene on the chromosome 1 in the q22–q23 region and in mice, USF-1 was assigned to chromosome 1 (Henrion et al., 1995). The human and murine USF-2 genes were assigned to chromosomes 19q13 and 7 respectively (Groenen et al., 1996).

Protein structure and protein interaction

The cloning of USF-1 and USF-2 revealed that both proteins are members of the highly conserved family of basic-Helix-Loop-Helix-Leucine Zipper proteins (bHLH-LZ) (Figure 1C) (Atchley and Fitch, 1997). These include Myc (Murre et al., 1989) and the Myc-subfamily factors Max (Blackwood and Eisenman, 1991), Mad (Ayer et al., 1993) and Mxi1 (Zervos et al., 1993) as well as the microphthalmia-associated transcription factor, Mitf (Tassabehji et al., 1994), and the highly-related factors TFEB (Carr and Sharp, 1990), TFEC (Yasumoto and Shibahara, 1997), TFE3 (Beckmann et al., 1990).

The bHLH-LZ proteins have a highly conserved DNA-binding domain, composed of a basic (b) region, followed by a helix-loop-helix (HLH) and a leucine zipper (LZ) motif (Figure 1B). Thus, the USF-1 and USF-2, b-HLH-LZ regions have 70% identity while the full length proteins present only 44% identity (Sirito et al., 1994). Only limited regions of sequence homology could be found in the USF-1 and USF-2 N-terminal regions, one of which is highly conserved and located upstream from the basic region. This USF-specific region named USR and encoded by the exon 6, has been shown to be essential for transcription activation (Groenen et al., 1996; Luo and Sawadogo, 1996; Qyang et al., 1999) although how it operates is unknown. The presence of the contiguous HLH and LZ domains distinguishes these proteins from bHLH proteins (e.g. MyoD family, E2A) (Blackwell and Weintraub, 1990; Kophengnavong et al., 2000), bZip transcription factors (e.g. c-jun and c-fos) (Busch and Sassone-Corsi, 1990) and from other related-transcription factors lacking a basic region that act as negative regulators and does not bind DNA (e.g. Id) (Benezra et al., 1990; Ellis et al., 1990).

The basic region is involved in DNA interaction with the E-Box consensus sites CANNTG (Baxevanis and Vinson, 1993), whereas the HLH and LZ domains are mainly involved in dimerization, even though LZ integrity is essential for specific and high affinity binding (Sha et al., 1995). The HLH and LZ amphipathic α-helices therefore together mediate selective pairing for dimer formation. Consequently, protein interactions were thought to be limited to members of each b-HLH and b-HLH-LZ family subgroup (Baxevanis and Vinson, 1993; Ciarapica et al., 2003) because of recognition specificity. Therefore, USF-1 and USF-2, were first shown to interact only with each other, leading to homo- and heterodimer formation (Sirito et al., 1992, 1994), an observation consistent with the conserved region being limited to the C-terminal region that encompasses the dimerization and DNA-binding domains. More recently, USF-1 has been reported to associate in vivo with Fra1, a member of the b-Zip protein family, to promote transcription, demonstrating that cross-talk occurs between distant members of the protein family (Pognonec et al., 1997). The Cha bHLH transcription factor has also been shown to interact in vivo and in vitro with USF-1. The Cha/USF-1 dimer can bind an E-box motif, inhibiting USF-1 promoter activity and thereby negatively regulating USF-dependent transcription (Rodriguez et al., 2003). USF-dependent gene expression is also negatively regulated through interaction with the USF-2 splice variant, termed USF-2b (Figure 1A) (Howcroft et al., 1999; Viollet et al., 1996). The USF-2b variant lacks exon 4, which contains an additional positive-regulatory domain, whereas the USR and the negative-regulatory domain remain functional. The USF-2b variant forms heterodimeric complexes with the USF proteins, which can bind the E-box motif but fails to promote gene activation. The USF-2b variant is thus a dominant negative regulator of USF-dependant gene expression. Similarly, alternative splice events also affect USF-1 mRNA (Gao et al., 1997; Gregor et al., 1990; Saito et al., 2003). The excision of a part of exon 4 generates a novel USF-1 isoform, resulting in new heterodimeric complexes that can modulate USF-dependent gene regulation (Saito et al., 2003).

DNA-binding properties

Symmetrical E-box motifs are the principle target-binding sites for members of the b-HLH and b-HLH-LZ transcription factor families. The canonical E-box sequence is only six nucleotides long, with some potential degenerate bases. Although its short sequence favours a broad distribution over the genome, only a restricted set of E-box motifs have been identified as DNA control elements. Sequential selection and amplification of binding-sites analysis (Blackwell and Weintraub, 1990; Blackwell et al., 1993) and DNA-binding assays (Aksan and Goding, 1998; Dang et al., 1992; Galibert et al., 1993; Halazonetis and Kandil, 1991) have been used to investigate the specific protein-DNA binding that discriminates one b-HLH-LZ transcription factor from another. Taken together, these results show that the CANNTG core E-box sequence is usually conserved, that the two central nucleotides (NN) are in most cases either GC or CG, and that altering the spacing between the CA and TG abrogates DNA binding. Conservation of the short palindromic sequence is therefore important, even if one nucleotide variation is tolerated. For example, USF-1 has a higher binding affinity for the CACGTG core motif than for the CATGTG motif that is also recognized in vivo by Mitf, the ‘master regulator’ of melanocyte development. Consistent with these observations, the crystal structure resolution of b-HLH-LZ proteins bound to cognate DNA reveals critical and specific contact points between nucleotides and amino acids in the basic region (Ferre-D'Amare et al., 1994). Importantly, the nucleotides flanking the CANNTG core sequence may also affect significantly the selectivity of the different E-box-binding proteins. A 5′-T residue is critical for Mitf binding to CATGTG motifs (Aksan and Goding, 1998), whereas it prevents Myc binding to CACGTG elements (Fisher et al., 1993). Similarly, the binding sites of several USF-1 regulated promoters, including adenovirus major late (AdML) (Sawadogo and Roeder, 1985), haeme oxygenase (Sato et al., 1989), metallothionein I (Andrews et al., 2001), adenomatous polyposis coli (APC) (Jaiswal and Narayan, 2001), DNA topoisomerase IIIα (Han et al., 2001), deoxycytidine kinase (Ge et al., 2003) among others possess AC residues 3′ to the core E-box motif. These in vivo observations suggest that USF-1 can selectively recognize the CACGTGAC sequence. Thus requirement for particular residues flanking the core E-box may provide a mechanism in vivo to recruit only specific subsets of E-box-binding factors to specific subsets of E-box elements. Notwithstanding these observations on sequence specificity, in some contexts USF-transcription factors can bind and activate transcription from degenerate E-box sequences where no doubt promoter context dictates potential protein–protein interactions, which in turn might modulate binding of E-box factors.

Upstream stimulating factor-1 binding to specific E-box motifs can also be modulated depending on the cell type and context. Methylation at the CpG site located centrally within the E-box motif (CACpGTG), strongly inhibits complex formation (Prendergast and Ziff, 1991; Watt and Molloy, 1988) and regulates negatively gene expression. This has been described, among others (Griswold and Kim, 2001; Kumari and Usdin, 2001), for the metallothionein-I gene (Majumder et al., 1999), leading to gene silencing in mouse lymphosarcoma cells. Single nucleotide polymorphisms (SNPs), within the E-box core motif, also modulates gene regulation. For example, a single G/C base transition within the USF E-box consensus of the thymidylate synthase gene, implicated in folate metabolism, prevents USF complexes from binding to its cognate sequence (Mandola et al., 2003). Considering the high frequency of the G/C polymorphism in all major racial and ethnic groups, the knowledge of this SNP would be extremely useful to tailor individual 5-fluorouracil-chemotherapy.

Multiple signal transduction pathways can also modulate the DNA-binding activity of the USF-1 protein. These pathways include the p38 stress activated kinase pathway (Galibert et al., 2001), the protein kinase A and C pathways (Xiao et al., 2002), the cdk1 pathway (Cheung et al., 1999) and the PI3Kinase pathway (Nowak et al., 2005). The p38, pKA, pKC, cdk1 kinases all phosphorylate USF-1, increasing the binding of this factor to its E-box-containing target genes. However, the PI3Kinase-regulated form of USF-1 does not bind to the E-box motif. Taken together, these DNA and protein modifications would allow precise modulation of USF gene-dependent regulation.

USF expression

Although the expression of the USF-1 and USF-2 proteins is ubiquitous (Gregor et al., 1990; Sirito et al., 1992, 1994), distinct cell-specific ratios of USF homodimers and heterodimers are found in different cell types (Sirito et al., 1994; Viollet et al., 1996). Even though, USF dimers have similar DNA-binding properties they may control differently target genes, by establishing specific interactions with different transcription factors through their distinct N-terminal domains (Kirschbaum et al., 1992; Luo and Sawadogo, 1996). Thus the cellular functions of the different USF-family members may not be completely redundant. Accordingly, USF-1 and USF-2 knock-out mice display distinct phenotypes, including an asymmetric influence of each gene on the expression of the other (Casado et al., 1999; Sirito et al., 1998).

USF factors are key components of the transcription machinery

The USF proteins participate in the transcriptional activation of many genes implicated in a variety of functions and networks through distinct hierarchical mechanisms. USF transcriptional activity is dependent on the integrity of the b-HLH-LZ domains for efficient DNA binding and dimerization and on the presence of activation domains located in the N-terminal region of the proteins (Figure 1B). First, USF proteins have been found to modulate gene transcription through their binding to cognate E-box motifs leading to transcription stimulation (Ferre-D'Amare et al., 1994). Secondly, interaction between USF-1 and general and cell-specific transcription factors SP1, Pea3 and MTF1, respectively, for example, leads to cooperative transcriptional regulation (Figure 2) (Andrews et al., 2001; Ge et al., 2003; Liu et al., 2004). The USF-1/Pea3 ternary complex is dependant on the presence of the LZ motif, with Pea3 recruitment by the LZ motif leading to cooperative transcription of the Bax gene (Firlej et al., 2005). Thirdly, USF-1 interacts directly with the transcriptional machinery of TATA-plus and TATA-less promoters. TFIID and TBP associated factors involved in the pre-initiation complex of TATA-directed gene transcription, do interact with USF-1 (Figure 2) (Bungert et al., 1992; Chiang and Roeder, 1995; Meisterernst et al., 1990; Sawadogo and Roeder, 1985; Sawadogo et al., 1988). USF-1 has also been shown to bind to the pyrimidine-rich initiator (Inr) element located near to the start site in TATA-less promoters (Du et al., 1993; Roy et al., 1991; Smale and Baltimore, 1989). Similarly, the TFII-I factor, a potential-binding element of the Inr sequence that stimulates transcription from TATA-less and Inr-containing promoters, is also able to bind to a distinct E-box element of the AdML promoter, previously identified as the USF recognition site (Roy et al., 1991, 1993). The interaction of USF-1 and TFII-I at E-box and Inr elements, is consistent with USF-1 protein structure data (Roy et al., 1997). Indeed, hydrodynamic measurements of USF protein–DNA complex showed that USF factors can exist as a bivalent homotetramer, which is dependent on the integrity of the LZ domain (Ferre-D'Amare et al., 1994). Homotetramers could potentially interact simultaneously with two DNA recognition sites to allow DNA looping that would modulate gene transcription regulation (Sha et al., 1995). Finally, it has been shown that USF-1 mediates recruitment of enzymes, such as PCAF that acetylates histones, and SET7/92 that methylates histone H3K4 (West et al., 2004) (Figure 2). These recruitments allow chromatin remodelling and opening, promoting DNA loading of the transcription machinery and transcription activation. Accordingly, USF-1 interacts preferentially with highly acetylated histone H4 nucleosomal DNA (Vettese-Dadey et al., 1996). Furthermore, USF proteins have been found to direct topoisomerase III (hTOP3α) gene expression, which catalyses changes in the topological form of DNA (Han et al., 2001). Modification of DNA supercoil structure allows chromatin accessibility and modulation of gene expression.

Figure 2.

Upstream stimulating factor-1 transcriptional cooperation.

Taken together, these data describe USF-1 as a crucial and general transcription factor having multiple roles during transcription regulation, consistent with its ubiquitous expression and broad distribution of its cognate E-box motif across the genome. Consistent with this, the number of USF-1-dependent genes is extremely high with some playing a critical role during development and in the adult. Indeed, USF-1/USF-2 double knock-out mice display an embryonic-lethal phenotype, suggesting that a minimal USF activity is essential for embryonic development.

USF-1 a stress-responsive factor

USF-1 and the tanning response in melanocytes

Skin cells are the first line of defence protecting the organism against a variety of chemical, physical and biological hazards. Solar ultraviolet radiation is a major environmental hazard and can cause skin inflammation, ageing and oncogenesis. Melanocytes protect against UV-mediated DNA damage through the tanning response in which UV irradiation triggers melanocyte production of melanin. Melanin synthesis takes place in specific organelles termed melanosomes that are transferred to neighbouring keratinocytes, thereby generating a protective skin screen. Melanin production is the result of a complex signalling network involving paracrine factors secreted by the keratinocytes [α-melanocyte specific hormone (MSH) encoded by the POMC gene, endothelin and FGF-2], specific receptors (melanocortin 1 receptor MC1R, endothlin B receptor EDNRB, tyrosine kinase receptors respectively) and transcription factors (Mitf, USF-1) as well as genes directly involved in pigment manufacture (Tyrosinase, TRP-1 and Dct) (Hirobe, 2005; Kadekaro et al., 2003). Although the Mitf transcription factor is required for constitutive pigmentation (Busca and Ballotti, 2000), the p38-stress activated USF-1 transcription factor is necessary for the tanning response. Indeed, both are member of the b-HLH-LZ family and bind to E-box regulatory elements, and both are target proteins of the stress-activated p38 kinase (Galibert et al., 2001; Mansky et al., 2002; Vance and Goding, 2004), leading to the idea that Usf-1 and Mitf may act in concert or sequentially to activate UV-responsive genes (Vance and Goding, 2004). However, USF-1 is a critical UV-responsive activator of POMC, MC1R, Tyrosinase, TYRP-1 and Dct genes (Figure 3). Indeed, these genes failed to be activated following UV stimulation in USF1 −/− melanocyte cell line (Corre et al., 2004). Moreover, although endothelin and FGF-2 UV-induced expression is USF-1 independent, they both activate p38 kinase in synergy with α-MSH, leading to a potential increase in USF-1 activity. The presence of this positive-feedback loop in the tanning response to UV-irradiation participates in the accumulation of protective pigment (Figure 3).

Figure 3.

Upstream stimulating factor-1 and UV-induced pigmentation. In response to UV, activation of the p38 kinase allows USF-1 phosphorylation on threonine 153. Binding of the phosphorylated-USF-1 protein to specific-E-box motifs on tyrosinase, POMC and MC1R promoter genes induce gene expression.

USF-1 and the immune response

The identification of USF-1 as a stress-responsive transcription factor is important in understanding the global transcriptional response to cellular stress. An increasing number of USF-1-regulated genes are implicated in the immune response following viral and bacterial infections. Two broad classes of immune response in which USF-1 participates can be described: the humoral-antibody response mediated by B-cell production of immunoglobulin (Ig), and the cell-mediated immune response involving the cytotoxic and helper T-lymphocytes (Figure 4).

Figure 4.

Upstream stimulating factor and the immune response.

E-box motifs are critical regulatory elements within the enhancers and promoters of Ig chain genes (Ephrussi et al., 1985). USF-1 has thus been shown to bind and promote transcription of the λ2-Ig light chain gene (Chang et al., 1992), while TFE3 is thought to bind the μE3 E-box within the Ig heavy-chain enhancer gene (Carter et al., 1997). Interestingly, a 50-bp repeat (VNTR) within the D/J interval of the Ig heavy chain locus presents an E-box-binding element. This VNTR, which is entirely deleted during D/J rearrangement in B-cell differentiation, is thought to interact and sequester transcription factors, potentially USF, before B-cell differentiation and Ig-gene expression (Trepicchio and Krontiris, 1993). During the primary immune response, IgM pentamer formation is dependent on the synthesis of the Ig J chain, which links the IgM monomers. Efficient J chain transcription in activated B cells is dependent on USF factors in combination with other regulatory elements (Wallin et al., 1999). Polymeric IgA and IgM antibodies are transported across mucosal and glandular epithelia by polymeric Ig receptors (pIgRs) to mediate the host defence. One molecule of pIgR must be synthesized for each molecule of Ig. Therefore, the expression of pIgR must be strictly regulated during B-cell activation, and has been shown to be dependent on USF proteins. Indeed, chromatin immunoprecipitation shows that USF proteins interact in vivo with the pIgR gene promoter (Bruno et al., 2004; Martin et al., 2003).

Circulating Ig binds specifically to foreign antigens targeting them for destruction either by phagocytosis or by activating the complement system. USF-1 participates in this immune response phase by regulating the C4 complement component gene expression (Galibert et al., 1997). USF-1 also regulates β2-microglobulin transactivation, which encodes a glycoprotein associated with the major histocompatibility class I complex molecules that play a central role in antigen presentation and in Ig transport (Gobin et al., 2003; Howcroft et al., 1999). USF-1 binds and activates the CIITA gene promoter, which is required by MHC class II gene induction by interferon-gamma, involved in cellular immunity (Muhlethaler-Mottet et al., 1998). Together, these data clearly show that USF-1 is a key regulator of the immune response.

Given the important role of USF factors in the immune response it is perhaps not surprising that a variety of pathogens make use of USF for their own ends. For example, Chlamydia, which are obligate intracellular pathogens, can escape T lymphocyte immune recognition by degrading the host USF-1 transcription factor required for class I and II MHC antigen expression in the infected cells (Zhong et al., 1999, 2001). Viruses have also evolved to manipulate and divert to their own benefit host responses following infection and use USF-1 for key roles during the virus replication cycle. Thus, USF is involved in transcriptional regulation of the AdML promoter (Gregor et al., 1990), of the HIV LTR (Giacca et al., 1992; Maekawa et al., 1991; Sieweke et al., 1998), Varicella-zoster virus (Meier et al., 1994) and Epstein–Barr virus (Liu et al., 1996), perhaps taking advantage of the USF as a factor activated by the kind of stress that might be triggered by viral infection. Furthermore, data suggest that USF-1 interacts also with the transcriptional virus machinery to enhance transcription at the E-box motif. Cooperation of USF-1 with the varicella-zoster virus immediate–early protein 62 (IE62) requires the DNA-binding domain and transcriptional activation domains of USF (Meier et al., 1994). Activation of the cellular p38 stress signalling pathway following viral infection (Zachos et al., 1999) may lead to efficient activation of USF factors which are then recruited for viral gene expression. Therefore, USF proteins are versatile transcription factors allowing both immune responses and pathogens proliferation.

USF and the cell cycle

Regulation of the cell cycle is a critical process that allows the correct timing and order of events prior to cell division. The progression of the cell cycle is thus strictly regulated to maintain genetic integrity and to ensure that genetic information is correctly passed on to daughter cells. Cyclin-dependent kinases (Cdks) and their activating subunits, the cyclins, are the driving force allowing cell cycle progression. During the cell cycle, kinase levels change as a result of a combination of gene transcription and subsequent proteolysis by the ubiquitin-proteasome system leading to regulation of S and M phases by different subsets of cyclin-cdk complexes. Regulation of the G1/S progression-transition is dependent on cyclins D and E in combination with Cdk2, Cdk4 and Cdk6, whereas the G2/M progression–transition is dependent on cyclins A and B in combination with Cdk1 (cdc2) (Hunt, 1991). The cyclin-cdk complexes are crucial to the continued and timely progression through the cell-cycle and are regulated by various mechanisms, including Cdks inhibitors (Peter and Herskowitz, 1994). USF transcription factors are involved at both the G1/S and G2/M transitions by regulating the expression of diverse cyclin and Cdk genes (Cogswell et al., 1995; North et al., 1999; Pawar et al., 2004; Philipp et al., 1994) (Figure 5). For example, CDK4 is a direct target of both USF-2 and the c-Myc b-HLH-LZ proteins and the evidence suggests that these two transcription factors bind simultaneously to distinct and adjacent E-box motifs within the CDK4 promoter, cooperatively activating transcription in non-tumourigenic breast epithelial cells (Pawar et al., 2004). In some breast cancer cell lines, where c-Myc is expressed at high levels and USF is transcriptionally inactive, CDK-4 expression has been shown then to be independent of both USF and c-Myc transcription factors.

Figure 5.

Upstream stimulating factor and cell proliferation.

The USF-1 transcription factor is also found to be a positive regulator of cyclin B1 and Cdk-1 gene expression (Cogswell et al., 1995; North et al., 1999). Cyclin B1 is expressed between the late S and G2 phases of the cell cycle with Cdk-1 being activated by phosphorylation and by sufficient cyclin B proteins being synthesized, promoting entry into the M phase. Phosphorylation of recombinant USF-1 by active cyclin-cdk complexes (cyclinA2-cdk1 and cyclinB1-cdk1) has been shown to increase its DNA-binding activity (Cheung et al., 1999), which suggests a positive amplification loop favouring the G2/M transition.

Antiproliferative properties of the USF family on transformed cells

The involvement of the USF family in cellular growth control has been studied using focus and colony-formation assays (Aperlo et al., 1996; Luo and Sawadogo, 1996) in which the USF family was shown to inhibit the transformation of primary rat cells mediated by c-Myc and Ras. This highlights the fact that although USF and Myc proteins share homologous bHLH-LZ domains, they have opposite roles in cellular transformation. In addition, only USF-2 was able to abolish E1A and Ras, or mutant p53 and Ras-induced focus formation of rat primary cells, demonstrating that despite their similarities, USF-1 and USF-2 does not possess redundant functions in regards with these genes (Luo and Sawadogo, 1996). The molecular process underlying the anti-proliferative properties of the USF proteins was shown to be dependent on USF-DNA binding activity, allowing interaction and regulation of target promoters. This is consistent with the fact that some USF-target genes are involved in the control of cellular proliferation and differentiation (Figure 5); USF transcription factors participate in the regulation of cyclin-cdk encoding genes, tumour suppressor genes, and growth factor networks (Kingsley-Kallesen et al., 1999; Szentirmay et al., 2003). Although the integrity of the USF DNA-binding domain is necessary to confer its anti-proliferative properties, it is not sufficient as they also require full transcriptional activity. In this respect, over expression of recombinant-USF proteins in HeLa cells causes marked growth inhibition, whereas no effect was observed in Saos-2 cells. Analysis of USF activity in both cell lines showed that the Saos-2 cells lacked transcriptionally active USF-1 and USF-2 forms (Qyang et al., 1999). Thus the transcriptional activity of USF proteins appears to be context dependent and may indicate that modulation of USF activity by transcriptional co-activators or post-translational modifications may be cell-type specific. If the USF family is indeed involved in the control of cell proliferation, de-regulation of USF family members should occur in cancer cells, contributing to uncontrolled proliferation. In support of this notion, loss of USF transcriptional activity has been observed in three of six transformed breast cell lines (Ismail et al., 1999) highlighting the potential of importance of USF family in cell proliferation control.

Expression of the human telomerase reverse transcriptase (hTERT) gene, which is necessary to avoid telomere shortening and cell crisis, is a major determinant of cellular longevity. Upregulation of hTERT transcription is observed in more than 85% of tumour cells, whereas its expression is shut down in most somatic cells. Several transcription factors are thought to regulate TERT expression and activation of TERT expression by b-HLH-LZ proteins involving Myc and USF subfamily members has been demonstrated (Goueli and Janknecht, 2003; Horikawa et al., 2002; Yago et al., 2002). The effect of USF proteins on hTERT regulation in tumour cells has been studied using hTERT-positive and hTERT-negative tumour cell lines (Goueli and Janknecht, 2003). USF-1 and USF-2 heterodimers interact in vivo and in vitro with E-box motifs within the hTERT promoter and USF proteins bind to promoters in both hTERT-positive and negative cells. However, stimulation of hTERT gene expression is observed only in hTERT-positive cell lines. In non-immortalized hTERT-negative cells, USF binding to the hTERT promoter is transcriptionally neutral. Therefore, on their own, USF proteins are unable to re-activate hTERT expression in hTERT-negative cells. Precisely how the feedback between USF and hTERT occurs is currently unknown, but through hTERT transactivation, USF proteins can, depending on the cell context, contribute to the maintenance of cell immortality.

USF and tumour suppressor genes

Importantly, given its proposed role in regulation of hTERT and in regulation of proliferation and transformation, distinct tumour suppressor genes, including p53, BRCA2 and APC, have also been shown to be direct targets of the USF transcription factors (Figure 5) (Davis et al., 1999; Hale and Braithwaite, 1995; Jaiswal and Narayan, 2001; Reisman and Rotter, 1993). The p53 tumour suppressor gene functions as a transcription factor (Levine et al., 1991) and plays a major role in regulating the response of mammalian cells to stress and damage, partly through transcriptional activation of genes involved in cell cycle control, DNA repair and apoptosis. In response to myriad stresses (DNA damage, hypoxia, proliferative signals, etc.), it is well established that the p53 protein becomes stabilized, causing cells to undergo either cell cycle arrest or apoptosis. Concomitant to DNA-damage, stress activated USF-1 factor would participate in DNA-repair through p53 and BRCA2 gene regulation (Davis et al., 1999; Hale and Braithwaite, 1995; Reisman and Rotter, 1993). As loss of function of tumour suppressor genes participates in the development of carcinogenesis, loss or reduced activity of USF, either via abrogation of the protein/DNA complex through E-box modification (mutation or methylation) or alteration of the USF transcriptional activity (post-translational modification, protein/protein interaction) could also in principle contribute to transformation, by leading to reduced p53 expression for example.

USF and glucid–lipid metabolism

In addition to their multiple roles in regulating proliferation, an additional function of USF factors is in glucid and lipid metabolism (Figure 6), where they regulate a wide number of genes characterized by the presence of E-box motifs within their promoter regions. USF-1 participates, among others, in the regulation of the insulin gene (Read et al., 1993), the insulin growth factor-binding protein 1 gene (IGF-BP1) (Matsukawa et al., 2001), the glucagon receptor gene (Portois et al., 2002), the islet-specific glucose-6-phosphatase catalytic subunit-related gene (Martin et al., 2003), and the glucokinase gene, which are critical elements in hepatic glucose sensing (Iynedjian, 1998; Roth et al., 2004). Moreover, disruption of the genes in mice allowed the implication the USF factors in regulation of the L-Pyruvate Kinase (L-PK) gene. The hepatic USF binding activity in USF-1 and -2 KO mice is then accounted by the remaining USF-homodimer species (Vallet et al., 1997, 1998). USF-1 is also largely involved in lipogenesis, regulating among others fatty acid synthase, apolipoprotein, hepatic lipase, acetyl-CoA and carboxylase genes. Finally, familial combined hyperlipidaemia (FCHL), characterized by elevated levels of total cholesterol and/or triglycerides, was recently linked and associated to the USF-1 gene (Coon et al., 2005; Pajukanta et al., 2004). Additional studies (Elbein et al., 1999; Putt et al., 2004) suggest that the involvement of USF-1 in lipid and glucose metabolism may not be limited to FCHL and may be linked to other diseases such as type 2 diabetes.

Figure 6.

Upstream stimulating factor and glucid–lipid metabolisms.

Conclusions

Upstream stimulating factor proteins are stress-responsive b-HLH-LZ factors that regulate a wide range of target promoters through specific-E-box motifs. Identification of the repertoire of cellular genes that are regulated by USF has been complicated by the fact that E-box motifs are widely represented across the genome and because the majority of E-box-binding sites can also be recognized by other b-HLH-LZ members, including the Myc and Mitf families. Nevertheless, the number of genes known to be regulated by USF transcription factors is constantly expanding while it has become clear that USF DNA-binding and activity is subject to regulation through several pathways, most notably by stress-activated signalling. Despite these advances several key questions remain to be answered. For example, do the various post-translational modifications of USF leads to exchange of co-factors used by USF to mediate gene regulation, and indeed what is the full repertoire of co-factors recruited by USF proteins to target promoters? For genes where E-boxes are recognized by both USF and other transcription factors, is recognition sequential and which binds first? For example, it is possible that USF factors play a key role in maintaining a chromatin environment at a promoter that is critically required for other factors to regulate transcription. In addition, there is some evidence for apparently conflicting roles of USF as USF family members can either allow progression through the cell cycle by activating cyclin-cdk gene expression, or repress the cell cycle by regulating tumour suppressor genes; and if USF is an anti-proliferative factor, why does USF regulate hTERT gene expression in cancer cells? Most likely many aspects of USF function will be cell type and context dependent and indeed, like many other factors, it may even switch from activator to repressor depending on which signal transduction pathways are operating at a given time. Thus, despite the increasing evidence that USF plays a key role in many aspects of cell division and proliferation, there remains a great deal to learn.

Acknowledgements

We would like to thank Prof. M. Philippe and Prof. F. Galibert for critical reading of the manuscript. We apologize to those whose work was not cited because of space limitations.

Ancillary