The world according to MafHozumi Motohashi, Jordan A. Shavit1, Kazuhiko Igarashi, Masayuki Yamamoto and James Douglas Engel1,*
Institute of Basic Medical Sciences and Center for TARA, University of Tsukuba, Tsukuba 305, Japan and 1Department of Biochemistry, Molecular Biology and Cell Biology, Northwestern University, Evanston, IL 60208-3500, USA
Received April 11, 1997;Revised and Accepted June 11, 1997
ABSTRACT
Maf family proteins are so named because of their structural similarity to the founding member, the oncoprotein v-Maf. The small Maf proteins (MafF, MafG and MafK), as do all family members, include a characteristic basic region linked to a leucine zipper (b-Zip) domain which mediate DNA binding and subunit dimerization respectively. The small Maf proteins form homodimers or heterodimers with other b-Zip proteins present in the cell and bind to Maf recognition elements (MARE) in DNA. Since they lack known transcriptional activation domains, the small Maf proteins function either as obligatory heterodimeric partner molecules with numerous large subunits, discussed below, or alternatively as homo- or heterodimeric transcriptional repressors. The three small Maf proteins are expressed in a number of overlapping tissues, but their expression profiles nonetheless appear to be under meticulous tissue- and developmental stage-specific control. The MARE bears a striking resemblance to the NF-E2 binding sequence. NF-E2 binding sites in the human [beta]-globin locus control region have been directly implicated as integral components in the circuitry required for eliciting changes in chromatin structure that precede globin gene activation. While the NF-E2 DNA sequence has been shown to be important for erythroid-specific gene regulation, a growing list of other genes may also be regulated through the same, or very similar, cis elements in non-erythroid cells. Taken together, these observations argue that comprehensive analysis of the activities of the small Maf proteins may provide a unique perspective for expanding our understanding of transcriptional regulation that can be elicited through interacting transcription factor networks.
INTRODUCTION
Mammalian embryogenesis is a complex process during which initially naive, often multipotent, cells proliferate, migrate and differentiate in response to inductive cues to form the tissues that will eventually comprise an independent organism. During this time, numerous signaling and response genes are turned on and off in different developing and migrating cells, tissues and organs. This process is regulated by a variety of extrinsic and self-signaling events which ultimately lead to the nucleus. There the transduced signals exert regulatory control over transcription factors that bind to specific cis-regulatory elements which then activate or repress expression of specific sets of genes. In this regard, numerous oncoproteins and related cellular factors have become conspicuous as critical players in this intricate regulatory pavane (1 ).
The founding member of the Maf protein family (v-Maf) was originally discovered as the transduced transforming component of avian musculoaponeurotic fibrosarcoma virus, AS42 (2 ). Subsequent studies identified the cellular homolog of this gene, c-maf, from which the v-maf oncogene was originally transduced (3 ), but in addition, c-maf was found to be but one member of an extended multigene family. Products of the maf proto-oncogene and related family members (the Maf family proteins) share a common, relatively well-conserved basic region and leucine zipper (b-Zip) motif which mediate DNA binding and dimer formation. Members of the Maf family are divided into two subgroups: the large Maf proteins, c-Maf (3 ), MafB (52 ) and NRL (5 ), all of which contain a distinctive acidic domain that probably enables transcriptional activation, and the small Maf proteins, MafK (6 ), MafF (6 ) and MafG (7 ), all of which lack activation domains.
A direct physiological role(s) for Maf family proteins remained elusive from 1989, when c- and v-maf were originally described, until ~4 years later, when the small Maf proteins were first shown to function as one subunit of nuclear factor-erythroid 2 (NF-E2), an erythroid-specific transcription factor (8 ,9 ). NF-E2 was shown to be comprised of 45 and 18 kDa subunits, the former being p45 NF-E2 (the founding member of the vertebrate CNC transcription factor family; below) and the latter was shown to be a small Maf protein. Since then, studies detailing the functional activity of various large and small Maf proteins have appeared and reports describing their important contributions to development and differentiation have become more widespread (10 -13 ). In this review we summarize recent progress in analysis of the small Maf proteins (and, more briefly, describe the activity of the numerous and sundry partners of the small Mafs) and present speculation and evidence supporting the contention that both homodimers and heterodimers of the small Maf proteins exert finely articulated transcriptional control over gene expression during development, differentiation and oncogenesis.
THE CIS-REGULATORY TARGETS OF MAF PROTEINS
EXPRESSION PROFILES OF THE SMALL MAF PROTEINS
PARTNER MOLECULES OF THE SMALL MAF PROTEINS
The small Maf proteins form homodimers (7 ), heterodimers with one another (7 ) and also heterodimers with an ever-increasing constellation of additional b-Zip proteins (9 ,23 ,29 ,32 ,33 ). The first identified partner molecule of the small Mafs was p45, which was initially isolated as the large subunit of transcription factor NF-E2 (23 ,34 ,35 ). We and others then showed that the small Maf proteins serve as the lower molecular weight subunit (also referred to as the ubiquitous p18 subunit) of NF-E2 (8 ,9 ). Recently several p45-related molecules that participate in heterodimeric complex formation with the small Maf proteins have also been identified, named Nrf1, LCRF1 or TCF11 (36 ,37 ,60 ) and Nrf2 or ECH (32 ,38 ,39 ). Nrf1 and Nrf2 cDNAs were originally cloned from human erythroleukemia cells, whereas ECH was cloned from a chicken erythroid cell cDNA library. The structure of ECH is very similar to that of Nrf2, suggesting that the two are functional homologs. These three molecules (referred to as p45 NF-E2, Nrf1 and Nrf2) then form the `CNC family' proteins (36 ), since they share the conserved b-Zip structure originally identified in the Drosophila CNC (cap'n'collar) protein (40 ).
Expression of p45 is restricted within hematopoietic cells and intestinal epithelia, while the expression profiles of both Nrf1 and Nrf2 show somewhat broader distribution. Nrf1 is strongly expressed in heart and skeletal muscle, kidney, lung and ovary (37 ), whereas Nrf2 is most prominently expressed in kidney, lung, fetal liver and fetal as well as mature muscle (38 ). ECH mRNA was most abundant in peripheral blood and was induced during differentiation of chicken erythroid cell lines (32 ). As is also the case for p45, neither Nrf1 (29 ) nor Nrf2 (29 ,32 ) form homodimers that effectively bind the T-MARE DNA sequence. Instead, all three CNC family proteins appear to form obligate heterodimers with one or another of the small Maf proteins, which then allows them to bind to a MARE to activate transcription (9 ,29 ,30 ,41 ).
What are the possible advantages in this scheme of compulsory heterodimeric interactions between these varied (CNC and small Maf family members) b-Zip factors that only together constitute the final activator proteins? One possible answer is that subtle variations in DNA binding specificity or in trans-activation or trans-repression potential (below) may well be generated through dimer formation of only a limited number of b-Zip transcription factor partners. This would then theoretically result in combinatorially complex, but exquisitely sensitive control over gene expression. In this manner, the small Maf proteins enable CNC, as well as other b-Zip proteins, to bind to DNA and exert their function. Thus different biological activities, all elicited through sometimes slightly different MARE binding sites, may come into play depending on the partner molecules with which the small Maf proteins heterodimerize. The available CNC and small Maf proteins that are present in any given tissue or at different stages of differentiation in that tissue, as well as the relative affinities between them, probably determine the final quality and quantity of transcriptional effects exerted at individual MARE sites.
In this regard, it should be noted that the small Maf proteins do not possess a canonical trans-activation domain and homodimers of the small Maf proteins have been shown to act as direct transcriptional repressors. It has been demonstrated that regulation from MARE sites can be turned on and off in living cells by experimentally manipulating the balance between the small Maf proteins and partner molecules that contain trans-activation domains (Fig. 3 ). If the abundance of CNC partner molecules is inadequate to `titrate' all of the small Maf proteins produced in a cell, homodimers of the small Mafs would then be predicted to exert a dominant effect, leading to silencing through the MARE sites (9 ). We suspect that this mechanism could be operative in vivo as well as in cell culture, since the switch from activation to repression was found to occur within only a 4-fold difference in abundance of the small Maf protein (Nagai et al., in preparation).
CONTRIBUTION OF THE MAF NETWORK TO DEVELOPMENT
Recent data from gene targeting experiments provide some insight to help us understand the roles the small Maf proteins might play in development. mafK null mutant mice were found to be viable, fertile and apparently normal and healthy (49 ). Their peripheral blood cell counts, red cell parameters and blood smears were also within the normal range. NF-E2 binding activity in fetal livers of mafK null mutants was indistinguishable from that of the wild-type animals (49 ), suggesting the presence of a fully complementing activity. Since the same three small Maf proteins found in chickens (50 ) are also expressed in the mouse (unpublished observations), the most likely explanation is that MafF or MafG replace the ablated mafK gene function, since the chicken small Maf proteins are able to functionally substitute for one another (9 ).
Additional key insights into small Maf function were also provided from forced expression experiments (50 ,51 ). When mafK was stably transformed into murine erythroleukemia (MEL) cells under the control of a conditionally inducible (metallothionein gene) promoter, the MEL cells were induced to differentiate simply by the addition of zinc to the culture medium (i.e. in the absence of any other inducers; 50 ). This result suggested that quantitative control over mafK expression was important for the erythroid differentiation process (see Fig. 3 ). The result could be interpreted in two ways. The ectopic increase in MafK could have resulted in recruitment of a heterodimeric partner molecule which could then activate transcription of terminal erythroid genes through MARE sites. Alternatively, increased levels of MafK could have led to formation of homodimers which then repressed transcription from target genes required for proliferation and which normally inhibit differentiation of MEL cells. A likely resolution to these alternative explanations for the role of small Maf proteins in erythroid differentiation was suggested by preparing a dominant negative mutant of MafK (dnMafK), which was able to form homo- and heterodimers but rendered any such complex unable to bind to a MARE site (51 ). Expression of dnMafK in MEL cells lowered the overall binding activity to MARE sites in cell extracts, as anticipated, and led to decreased expression of the globin genes (51 ). These results suggested that under normal conditions MafK might be limiting in uninduced MEL cells and that its forced expression allowed complex formation with an activating partner molecule that was already present in the cell.
Recently several publications have appeared which underscore the developmental significance of the large Maf family proteins during neurogenesis. c-Maf was shown to be an important regulator of neuron-specific L7 gene expression in Purkinje cells (13 ). MafB plays an apparently critical role in segmentation of the hindbrain (11 ), as concluded from the discovery that the mouse Kreisler (kr) phenotype is due to mutation of the mafB gene. The neural retina is another site where large Maf family proteins play prominent roles. Both the opsin and rhodopsin genes were shown to be targets for the activity of NRL (27 ,28 ) and the retina-specific QR1 gene has recently been shown to be activated by both c-Maf and MafB (53 ). The expression profile of the small Maf protein MafK largely overlaps that of the large Maf proteins in neural tissues (31 ).
While hematopoiesis has been a focal topic for CNC family activity for some time, large Maf molecules are also emerging as equally important players in hematopoietic differentiation. c-Maf was found to be a key regulator of Th2 (one subset of CD4+ T helper cells)-specific expression of IL-4 (12 ). The cis elements targeted by c-Maf in the IL-4 promoter differ substantially from the MARE consensus sequence, which may explain the reason why synergy with NF-AT is critical for c-Maf to function on the IL-4 promoter. MafB was shown to interact with Ets-1 to inhibit trans-activation of the transferrin receptor gene and additionally was found to exert an inhibitory effect on differentiation of chicken erythroblast cell line HD3 (10 ). MafB is highly expressed in myelomonocytic cells, but not in erythroid cells. Thus MafB is a good candidate to play an important role in myelomonocytic cell differentiation, possibly by inhibiting expression of erythroid genes.
Gene disruption of the CNC family members has also been reported. A p45 null mutant mouse showed unexpectedly severe hemorrhage due to a lack of platelets, indicating that p45 is essential for megakaryopoiesis (54 ). Surprisingly, erythropoiesis was not affected significantly by the mutation, suggesting that activation signals for erythroid transcription can be transduced through MARE sites even in the absence of p45. Another member of the CNC family, Nrf2, has also been disrupted (55 ) and Nrf2 null mutant mice developed normally and were fertile. Nrf2 null mutant mice were not anemic, again indicating that erythropoiesis was not significantly affected by loss of Nrf2. Very recently the third member of the CNC family, Nrf1, was disrupted (56 ). While Nrf1 null mutant mice die prior to 7.5 d.p.c. from a failure to complete gastrulation, Nrf1 null ES cells injected into wild-type blastocysts contributed to all mesodermal lineages tested, including blood, indicating that the Nrf1 homozygous null mutation results in a defect in hematopoiesis that is not cell autonomous. These results provoke the question: which is the genuine partner molecule of the small Maf proteins with which it binds to and activates erythroid MARE sites? Is there a new molecule involved, or are p45, Nrf1 and Nrf2 mutually compensating for one anothers' loss in erythroid cells?
The [beta]-globin LCR is generally agreed to confer the ability to regulate transcription by first initiating structural alterations in chromatin (18 ,57 ); what it does thereafter is the subject of considerable current debate. In this regard, it is interesting to note that the Bach proteins contain a so-called BTB domain, which has been implicated in remodeling of chromatin structure (44 -46 ). We speculate that the Bach proteins are one of the important potential candidates for heterodimerizing partners of the small Mafs in erythroid cells, since the very real possibility exists that a heterodimer formed between a small Maf and a Bach protein could be a key regulator of LCR chromatin structure. Disruption of the Bach genes may provide further clues to this puzzle, since several of the predicted phenotypes are so clear.
Given this vast repertoire of potential partners for the small Mafs, studies are now underway to determine which molecule is the physiological partner in each of the exemplary physiological situations. Still unidentified partners may yet play a significant role in this evolving discovery process. Characterization of heterodimers of the small Mafs and partner molecules, especially with regard to the differences they might evoke in transcriptional activity of specific target sequences, in their responses to signals from outside the cell and their influences on chromatin structure, will help to decipher the mechanisms of gene regulation during these varied cellular processes.
CONCLUSIONS
Just as one gently folds back the petals of an intricately layered flower to finally reveal the intrinsic beauty of the whole, detailed examination at each level of discovery has led to an ever-increasing comprehension and appreciation of the complexity of regulatory control over vertebrate gene expression elicited by Maf family transcription factors. The number of proteins that participate in the Maf network continues to grow, both through the discovery of new partner molecules that heterodimerize with the Maf leucine zipper motif and through apparently non-canonical protein interaction domains. Where this will all eventually lead is unknown, but the recent intersection of the Maf network with the AP-1 and subsequently with the T3R/RAR networks leads to anticipation that the story will become even more intriguing before it is complete.
Analysis of small Maf proteins, along with analysis of NF-E2, have led to significant new insights into the mechanisms that mediate lineage determination and differentiation. We envisage that establishing or elaborating a cell lineage occurs through combinatorial interactions that precisely balance the activities of all the factors that participate in this elaborate protein interaction network. This hypothesis suggests that it is the sum of a combination of interactive transcription factor affinities (both for one another and for the target sites to which each binds in DNA) that finally dictates cellular responses which either prohibit or initiate differentiation and growth programs. This concept thus advocates the inverse view of the `master regulator' hypothesis, where it is thought that a single determinative protein dictates the fate of multipotent cells to a particular differentiated tissue or cell type.
Finally, it is also interesting to speculate how this finely balanced intracellular equilibrium might become unbalanced during oncogenesis. One simple hypothesis suggests that cell transformation may be the manifestation of perturbation of this b-Zip network (7 ,42 ). It is well known that forced expression of components of transcription factor AP-1 induces cell transformation (58 ,59 ). As discussed in this review, the possibility certainly exists that inappropriately expressed AP-1 components may sequester small Mafs or other members of the b-Zip network from their partner molecules or target sequences, so that these AP-1 components can inhibit the cellular differentiation process and provoke cell transformation. To elucidate how this b-Zip protein network contributes to cell differentiation processes and how its perturbation leads to cell transformation, it is important to identify and comprehensively categorize the factors interacting with the Maf family proteins, to examine their modes of DNA binding and to analyze the interactions among them.
ACKNOWLEDGEMENTS
We thank Drs Kim-Chew Lim and Makoto Nishizawa for illuminating discussions. This work was supported by research grants from the Ministry of Education, Science and Culture (H.M., K.I. and M.Y.), the Japanese Society for Promotion of Sciences (M.Y.), the Uehara Memorial Foundation (M.Y.), the Mochida Memorial Foundation for Medical and Pharmaceutical Research (K.I.), the Ciba-Geigy Foundation for the Promotion of Science (K.I.), an MSTP training grant to Northwestern University (GM 08152, J.A.S.) and the NIH (GM 28896, J.D.E.).
14 Stamatoyannopoulos,G. and Neinhuis,A.W. (1994) In Stamatoyannopoulos,G., Nienhuis,A.W., Majerus,P. and Varmus,H. (eds) The Molecular Basis of Blood Diseases. (2nd edn) W.B.Saunders Co., Philadelphia, PA, USA. Chapter 4, pp. 107-155.
