Characterization of the Murine mafF Gene*

Small Maf proteins are obligatory heterodimeric partner molecules of mammalian Cap’n’Collar proteins that together control a wide variety of eukaryotic genes. Although both MafK and MafG are expressed in overlapping but distinct tissue distribution patterns during embryonic development, the physiological consequences of loss-of-function mutations in either gene are modest. This suggested that compensation by the third small Maf protein, MafF, might be a major reason for such mild phenotypes and that further analysis of MafF might therefore provide important insights for understanding small Maf regulatory function(s). We therefore cloned, mapped, transcriptionally and developmentally characterized, and finally disrupted themafF gene. We show that murine mafF is transcriptionally regulated by three different promoters and is most abundantly expressed in the lung. The lacZ gene inserted into the mafF locus revealed prominent expression sites in the gut, lung, liver, outflow tract of the heart, cartilage, bone membrane, and skin but not in hematopoietic cells at any developmental stage. Homozygous mafF null mutant mice were born in a normal Mendelian ratio and displayed no obvious functional deficiencies, indicating that MafF activity may be dispensable even in tissues where the expression of other small Maf proteins is quite low.

How a single cell differentiates into a variety of different tissues and organs to form a coherent multicellular organism is the central question of developmental biology. Transcription factors have been shown both genetically and biochemically to be key mediators of cellular differentiation, because they bind to specific DNA sequences to activate specific target gene transcription and often appear to be direct targets of cellular responses to inductive queues from the environment. The c-Maf transcription factor, originally discovered as a retrovirally transduced oncogene (1), was the founding member of a family of six related proteins that have been identified in vertebrate organisms. All of the Maf family members possess a conserved basic region closely flanked by a heptad repeat motif (bZip), which mediates DNA binding and dimer formation, respectively. These proteins commonly recognize a specific palindromic sequence, TGCTGACTCAGCA or TGCTGACGTCAG-CA, called a MARE (maf recognition element (2, 3)) through which they potentiate tissue-specific gene transcription.
The Maf family is subdivided according to molecular size. The large Maf proteins, c-Maf, MafB, and NRL, have similar conserved sequence domains and also possess potent transactivation sequences (2,4,5). The small Maf proteins, MafF, MafG, and MafK, do not possess trans-activation activity, and indeed these three genes encode proteins with little more than the basic and heptad repeat motifs, which identify them as members of the Maf protein family (6,7).
The small Mafs can affect transcription either by forming heterodimers with other bZip factors or by forming homodimers and heterodimers among themselves. When two small Maf proteins homo-or heterodimerize, they bind to MAREs and negatively regulate transcription, either indirectly, by competitively inhibiting the binding of activating molecules, or directly, by repressing transcription from the basal apparatus (8).
Over the last several years, three groups of small Maf interacting molecules, which can form either productive or unproductive complexes, have been described (3,9). The most physiologically important of these partner molecules are probably the vertebrate CNC 1 transcription factors (p45, Nrf1, Nrf2, and Nrf3), so named originally for their sequence homology to the Drosophila Cap'n'Collar protein (10). In general, the small Maf proteins heterodimerize with CNC family members to generate strong transcriptional activators, whereas CNC homodimers will not bind to MARE sequences (e.g. 11). MafK and p45 form a heterodimer that was the founding member of this family of vertebrate transcription factors called NF-E2 (12)(13)(14). Although p45 was originally believed to be important for erythroid-specific gene transcription, it has subsequently been shown to have a far more profound effect on megakaryopoiesis (15). The second group of small Maf-interacting proteins are the Bach factors (Bach1 and Bach2), which recent experiments suggest may be more important for their affects on chromatin assembly rather than on transcription per se (16,17). Finally, the small Maf proteins as well as some of the large Mafs and CNC proteins can also form heterodimers with Fos and Jun family members (2,7,18,19).
Loss of function mutants for several of the CNC proteins have been generated by targeted mutagenesis in embryonic stem (ES) cells followed by the generation of germ line mutant mice. p45 mutant mice displayed impaired platelet formation but little disturbance of erythropoiesis (15). nrf1 germ line mutation conferred a noncell autonomous mesodermal embryonic lethal defect (20,21), whereas nrf2 null mutants were viable (22) but had impaired reaction to xenobiotics (23). Because all three of these CNC proteins are capable of activating transcription from MARE elements in tissue culture co-transfection assays, the lack of prominent erythroid phenotypes in any of the mutant mice indicated that these activities could be compensating for one another in erythroid cells. This conclusion in turn indicated that unraveling the transcriptional responses mediated by individual MARE-binding proteins was going to be complicated.
To clarify in vivo functions of this extensively interacting regulatory network, we previously described null mutations in both the murine mafG and mafK loci. These studies showed that the mafK loss of function resulted in no discernable phenotype in homozygous mutant animals (24,25), whereas mafG mutant mice displayed both hematopoietic and behavioral deficiencies (25). Analysis of the mafK-and mafG-directed expression patterns (following insertion of lacZ into both chromosomal loci) showed, to our surprise, that both factors were expressed in highly tissue-restricted spatial and temporal patterns during embryonic development. This analysis clearly contradicted our earlier perception that the small Maf proteins were ubiquitously expressed (13,26). However, the lack of severe phenotypes in these mutant mice implied that the pattern of expression of the small Maf proteins might be extensively overlapping, suggesting that one small Maf might be able to compensate for the loss of another during development. If this hypothesis was correct, it would be very difficult to ascribe specific functional roles to individual small Maf family members because the effects of mutation of one protein would be masked by the expression of another.
To complete the initial analysis of this family to determine the contribution(s) of the small Maf proteins to this growing regulatory network, we cloned and characterized the gene encoding the third small Maf protein, mafF. MafF, like MafK and MafG, is expressed in a highly tissue-restricted pattern during embryonic development and is also found in some tissues that overlap expression of the other small Mafs in pattern, but it is also unique to a subset of tissues where the other proteins do not significantly accumulate. Homozygous mafF mutant mice were viable and apparently normal and healthy. The data show that although MafF, like MafK and MafG, is expressed in a distinct pattern during development, its loss of function in vivo is probably fully compensated (3).

EXPERIMENTAL PROCEDURES
Cloning the Murine mafF Gene-The mafF gene was cloned from a 129/SvJ genomic DNA library in Fix vector (Stratagene). The chicken mafF cDNA clone (6) was used as a probe under low stringency screening conditions. Filter hybridization was carried out in 50% formamide, 5ϫ SSC (0.15 M NaCl and 0.015 M sodium citrate), 20 mM Tris-Cl, pH 7.6, 1ϫ Denhart's solution, 10% dextran sulfate, and 0.1% SDS at 42°C overnight. Filters were washed in 2 ϫ SSC, 0.1% SDS at 37°C for 1 h. Positive clones were purified by three additional plaque purifications. The three mafF phage clones were subcloned into the XbaI site of Bluescript II (Stratagene). DNA sequences were determined by automated analysis (ABI, model 373A). Sequences of the murine mafF and mafG clones are available from Genbank (accession numbers AB009694 and AB009693, respectively).
Rapid Amplification of cDNA Ends Assay-5Ј-Rapid amplification of cDNA ends (5Ј-RACE) assays were performed utilizing the Marathon cDNA amplification kit following the manufacturer's instructions (CLONTECH). The sequences of the gene-specific primers (mMafF-F2 and mMafF-F1 nested primers) are shown in Table I. RACE products were subcloned into pGEM-T (Promega), and their sequences were determined as described above.
Chromosomal Mapping of mafF and mafG-DNA was obtained from the ([C57Bl/6JEi ϫ SPRET/Ei]F1 ϫ SPRET/Ei) interspecific backcross panel generated by The Jackson Laboratory (Bar Harbor, Maine; Jackson BSS). For the mafF locus, a 1.0-kbp XbaI fragment 5Ј to the second exon was used to detect a TaqI restriction fragment length polymorphism (RFLP). The more intensely hybridizing Mus spretus and C57Bl/6J bands were 6.5 kbp and 1.9 kbp, respectively, with 0.9 and 8.0 kbp faintly hybridizing fragments. TaqI digestion of DNA from each of 94 N 2 progeny was performed, followed by electrophoresis and blotting to a Nylon membrane. Hybridization and washing were performed using standard conditions. For the mafG locus, a 5.3-kbp BamHI fragment that spanned the second and third exons was used to detect the TaqI RFLP. The M. spretus fragment was 4.0 kbp, whereas C57Bl/6J DNA showed two bands of 1.9 and 1.7 kbp. In addition, there was a nonpolymorphic 1.3-kbp hybridizing band.
The typing data for the segregation of these two polymorphisms were submitted to The Jackson Laboratory Backcross DNA panel mapping resource and compared with approximately 3000 loci that had already been mapped. Linkage was observed for mafF to markers on chromosome 15 and for mafG to the distal end of chromosome 11. 2 Construction of the mafF Targeting Vector-To generate the mafF ES cell targeting vector, a 7.8-kbp SpeI-BamHI genomic fragment and a 1.9-kbp SphI-XbaI fragment were used. This replacement strategy removes a 1.1-kbp BamHI-SphI fragment that includes the entire mafF coding sequence except for the first three amino acids.
In the first step of the cloning strategy, nuclear localization signal (nls)-lacZ was introduced 3Ј to the 7.8-kbp SpeI-BamHI fragment, producing plasmid p7.8SB-nls-lacZ. To confirm that no mutations were introduced during cloning, the insert was sequenced to ensure that the nls-lacZ gene would be translated in-frame from the mafF initiation codon. Simultaneously, a loxP-flanked MC1promoter-neomycin resistance gene-poly(A)-MC1promoter-thymidinekinase-poly(A) (called MC1 neoMC1tk (25)) cassette and a MC1promoter-diphtheria toxin (MC1-DT (27)) cassette were cloned 5Ј and 3Ј to the 1.9-kbp SphI-XbaI genomic fragment, producing ploxPMC1neoMC1tk-1.9SX-DT. To minimize the possibility of affecting nls-lacZ expression, the loxPMC1neoMC1tk cassette was introduced in reverse orientation with respect to mafF. In the second cloning step, the insert of p7.8SB-nls-lacZ was introduced into ploxPMC1neoMC1tk-1.9SX-DT, producing the final mafF targeting vector.
Generation of mafF Mutant Mice-The targeting vector was linearized and introduced into R1 ES cells (28) by electroporation. Recombined ES cells were selected for G418 resistance (215 g of active component/ml; Life Technologies, Inc.). After 8 days of selection, G418resistant colonies were picked and subjected to PCR screening. After confirming homologous recombination by Southern blot analysis, positive ES cell clones were aggregated with CD-1 eight-cell stage embryos. After overnight culture, the blastocysts were transferred into the uterus of pseudopregnant foster females. Fourteen highly chimeric 129/ Sv ϩcϩp J/CD-1 mice (Ͼ80%) male mice were born starting with four different ES cell clones.
To confirm germ line transmission, the chimeric mice were intercrossed to CD1 females. Three of the ten chimeric animals transmitted the mutation through the germ line, and each was derived from a different ES cell clone. F1 and F2 heterozygous mafF mutants were intercrossed to determine any effects on viability, and the resulting pups were genotyped by PCR. The results were the same for all three mafF mutant lines.
Neomycin Gene Excision by Cre Recombinase-We removed the MC1neoMC1tk cassette using Cre-loxP-mediated recombination, because the MC1 promoter 5Ј to the neomycin or thymidine kinase genes could, in principal, affect nls-lacZ expression from the endogenous mafF promoter. For this purpose, we injected a Cre expression plasmid into fertilized eggs recovered from mafF heterozygous mutant females (CD1 background (33)). The resulting pups were analyzed for the presence or absence of the MC1neoMC1tk cassette by Southern blotting (data not shown). One mouse (line 801) was found to have precisely excised the positive selection cassette after Cre treatment, and after one additional backcross, lacZ expression was also analyzed in this line.
RT-PCR and Southern Blot Analysis-For the detection of mRNA expression, we used real time PCR analysis using a probe labeled with a reporter fluorescent dye (TaqMan probe). It was demonstrated that this analysis provides a quantitative measurement for mRNA levels (29).
cDNAs were synthesized from the RNA samples using murine leukemia virus reverse transcriptase (Superscript II; Life Technologies, Inc.) and random hexamer oligonucleotide primers. TaqMan probes, primers, and cDNAs were added to the master mixture that contained all reagents for PCR (Perkin-Elmer). Standard curves for MafF, MafG, MafK, and ribosomal RNA were generated by serial dilution of cDNA derived from intestine RNA. Sequences of the probes and primers are described in Table I. The primers for detection of MafF cDNA were mMafF-F and mMafF-R, and the probe was mMafF-P. mMafG-F, mMafG-R (primers), and mMafG-P (probe) were used for MafG. mMafK-F, mMafK-R (primers), and mMafK-P (probe) were used for MafK. The ribosomal RNA primers and probe (TaqMan Ribosomal RNA control reagents) were purchased from Perkin-Elmer.
The relative increase in reporter fluorescent dye emission because of the progression of the PCR reaction was monitored in real time by the 7700 sequence detector (Applied Biosystems). Signals were analyzed by the sequence detector 1.6 program (Perkin-Elmer). PCR conditions were as follows: 1 cycle at 50°C for 2 min, 1 cycle at 95°C for 10 min, and 40 cycles at 95°C for 15 s, 60°C for 1 min.
For RT-PCR analysis of mafF mutants, another set of primers (mmfu1 and mmfd1; Table I)  For Southern blots, a 0.4-kbp XbaI-KpnI fragment located outside of the targeting vector was used as a probe. The hybridization and washing conditions were described previously (25). To confirm homologous recombination in ES cells, five different restriction enzyme digests using three different probes were employed for Southern analysis.
LacZ Staining-Noon of the day that vaginal plugs were detected is taken as e (embryonic day) 0.5 or 0.5 dpc (days post-coitus). The fixation and 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside staining conditions were described previously (25). Thin tissue sections marking MafF expression were counterstained with eosin, whereas those displaying the MafG and MafK patterns were counterstained with nuclear fast red. All three mafF mutant lines, including one in which the neo gene was removed by cre excision, were analyzed for lacZ expression. There were no expression differences detected among any of the mafF lines.

RESULTS
The mafF Gene Is Transcribed from Three Different Promoters-The murine mafF locus was cloned from a 129/SvJ genomic DNA library using a chicken MafF cDNA (6) and screening under low stringency conditions. Three recombinants were identified, and all three were found to overlap the mafF locus. After detailed structural analysis (below), the murine mafF gene exons II and III were found to comprise the entire coding region. Comparison of the deduced amino acid sequence with the chicken MafF protein revealed 84% overall identity and less homology to other small Maf proteins, suggesting that this was indeed the mouse counterpart of chicken mafF (Fig. 1).
To determine the transcriptional initiation site for MafF mRNA, 5Ј-RACE analysis was performed ("Experimental Pro-cedures"). Exon Ib was found using adult lung mRNA as the initial template, and two additional first exons called Ia and Ic (as well as Ib) were found when 16.5 dpc whole embryo mRNA was used as the initial reverse transcriptase substrate (Fig. 2). The proximal promoter sequence of mafF exon Ib shows extensive homology with the chicken mafF first exon. 3 We conclude that the murine mafF locus is organized with two coding exons, in similar fashion to the mafG and mafK genes, but that three different promoters are alternatively spliced to the common exon II to generate three MafF mRNAs that encode the same protein product.
Chromosomal Mapping of the mafF and mafG Genes-Interspecific backcross mapping was employed to map the chromosomal position of the mafF locus. A TaqI RFLP was used for screening an interspecific backcross panel generated by The Jackson Laboratory ("Experimental Procedures"). After Southern blotting of the DNA from 94 backcross progeny, the results were submitted to a data base search containing approximately 3000 previously typed markers from the same backcross, thereby placing the mafF locus on chromosome 15 (Fig. 3). At the same time, we also determined that the mafG locus was on mouse chromosome 11 (Fig. 3). Together with a prior report describing the position of the mafK locus on chromosome 5 (30), these data complete the chromosomal positioning of all three small maf genes. Neither of the positions corresponds to a murine locus or human disease phenotype that might be because of the loss of mafF or mafG function (below).
Targeting of the lacZ Gene to Replace mafF Coding Sequences-To generate mafF null mutant mice, a targeting vector was designed as depicted in Fig. 4A. Except for the first three amino acids, the entire coding region of mafF was deleted by substitution with an in-frame lacZ fusion bearing a nuclear localization sequence ("Experimental Procedures"). At the same time, we introduced tandemly repeated loxP sites into the targeting vector, surrounding the neomycin resistance/thymidine kinase selection cassette, anticipating later excision by Cre recombinase, because it has been shown that the introduction of foreign promoters into genomic loci can disrupt normal gene regulation (e.g. Ref. 31).
The targeting vector ("Experimental Procedures") was electroporated into R1 ES cells, and clones were positively selected by culturing in the presence of G418. Of 240 ES cell transformants, 11 were found to be homologous recombinants by PCR and Southern blotting (Fig. 4B). Several of these ES clones were then aggregated with CD1 8-cell stage embryos to generate chimeras (32), and then F1 mice were generated by intercrossing the chimeras to CD1 females. From 10 initial chimeric   (33)) that were established using the three different ES cell clones gave identical genetic and expression results (below). mafF Is Expressed in a Highly Restricted Pattern during Embryonic Development-To clarify possibly unique versus potentially compensated MafF function(s), we analyzed lacZ expression in heterozygous or homozygous gene-targeted mafF mutant embryos as a reflection of natural MafF mRNA temporal-and tissue-specific accumulation. We then compared this pattern to those generated by the mafK-and mafG/lacZ mutant embryos (25).
LacZ expression was detected in the egg cylinder of mafF homozygous mutant embryos as early as 6.5 dpc, in the boundary between the extraembryonic and embryonic regions (Fig.  5A). This pattern is clearly distinct from that conferred by the mafG gene, which is expressed strongest in the embryonic region, and mafK, which is far more abundant in extraembryonic tissue. At 8.5 dpc, mafF/lacZ was weakly expressed in the future gut, allantois, and yolk sac endoderm but more strongly in the ectoplacental cone (Fig. 5B). Once again this pattern is in contrast to mafG expression at the same stage, which is strongly expressed throughout the embryo proper, whereas mafK expression is strong in the yolk sac endoderm and ectoplacental cone (25). At 9.5 dpc, strong mafF expression was observed in the primordial gut (Fig. 5C) and presumptive fetal liver (not shown), in the floorplate of the myelencephalon as well as in neural crest cells (not shown), and in spongiotrophoblasts and giant cells of the placenta (Fig. 5C). Although mafFdirected ␤-galactosidase synthesis was detected in embryonic hematopoietic organs, at 9.5 dpc, mafF/lacZ-expressing cells in the yolk sac were clearly endodermal rather than hematopoietic (Fig. 5D), whereas lacZ-expressing cells in the 12.5 dpc fetal liver appeared to be hepatocytes rather than the more rounded hematopoietic cells (Fig. 5E). These data suggested that MafF might not play a significant role in embryonic hematopoietic gene regulation, in contrast to MafK and MafG, which are both strongly expressed in hematopoietic cells (25). At 12.5 dpc, strong lacZ expression was also detected in the outflow tract of the heart (Fig. 5F) in a specific subset of dorsal root ganglia (Fig. 5G) and cranial nerve ganglia (not shown) as well as in the lung primordium and in the epithelium of the expiratory tract (below). Thus all three small maf gene expression patterns are quite distinct, but often overlapping, during embryogenesis.
In newborn animals, strong mafF expression was observed in keratinocytes (Fig. 6A) and in the cartilage and bone membrane (Fig. 6B). Intense expression persists in bronchial epithelia (Fig. 6C). As in earlier periods during development, both the bone marrow (Fig. 6D) and spleen (not shown) display scant  ␤-galactosidase staining, underscoring the possibility that mafF may not play a significant role in hematopoiesis.
Finally, we examined expression of all three small Maf mRNAs using a quantitative fluorescence ("Taqman") PCR assay. After preparing total RNA from various tissues in which we had previously detected expression of one or another of these mRNAs (or of the ␤-galactosidase reporter gene), we found that the primers used for PCR amplification of the three cDNAs interfered (to some extent) both with one another and with an internal control included in the reactions. We therefore quantified each mRNA relative only to its own peak expression level, and these results are presented in Fig. 7. Not unexpectedly, each of the small Maf mRNAs displayed a unique expression pattern: although MafK and MafF were most abundantly expressed in the lung, MafG appeared to be most abundant in the heart. Although MafG abundance differs by no more than 10-fold in any of the tissues examined, MafF abundance in different tissues varied by more than 65-fold.
The important conclusion that we can draw from this very sensitive assay is that one can detect each small Maf mRNA in each organ system or individual tissue that was analyzed. Whether or not this expression is representative of the multiple different cell types that constitute each organ system or whether in some cases low level expression might simply represent the level of contamination (e.g. of residual blood in muscle or heart) is not known at the present time. However, these data, when taken together with the lacZ expression data, suggest that although mafF might play important transcriptional regulatory roles during embryonic development, its loss of function could be masked in many, if not all, developing organs by compensation from the mafK and/or mafG genes.
mafF Null Mutant Mice Are Viable and Fertile-Homozygous mutant mice were generated by intercrossing CD1 F1 and subsequent generation mafF heterozygous mutant animals. The genotype of the pups was confirmed by PCR and Southern blotting (Fig. 4C). As anticipated, embryonic expression anal- . N, PCR-negative clone C, Southern blot analysis of wild type, heterozygous, and homozygous mafF mutant mice (line 2D12). DNAs were digested with BamHI. WT, wild type; Mut, mutant. D, RT-PCR analysis of total RNAs recovered from wild type, heterozygous, or homozygous mafF mutant 12.5-dpc embryos. MafF primers, located in exon II and exon III (Table I) ysis demonstrated that the deletion of virtually the entire coding region generated a null mutation in mafF. The complete absence of MafF mRNA in the homozygous mutant mice was confirmed by RT-PCR (Fig. 4D). As summarized in Table II, all possible genotypes of heterozygous mutant-intercrossed mice were born in the ratio predicted by Mendel's second law. Homozygous mutant mice appeared in all respects to be completely normal when compared with heterozygous or wild type littermates and showed normal body weight gain (data not shown) and fertility. Thus the loss of mafF activity results in no overt phenotype in animals.
mafF Homozygous Mutant Mice Display No Hematopoietic-, Skin-, or Bone-forming Disorders-As MafF is presumed to require heterodimerization with CNC family partner molecules to elicit transcriptional activation and several of these partner molecules have previously been implicated as key regulators in hematopoietic differentiation, we performed hematologic analysis on mafF null mutant mice. No decrease in the number of red or white blood cells or platelets in the hematocrit or in hemoglobin content was observed in homozygous mutant mice (Table III). This result is therefore consistent with the lack of significant MafF expression in hematopoietic cells. Because MafF is strongly expressed in bone membranes, we also analyzed the skeletal structure of the homozygous mutants by dissolving the tissue followed by alizarin red staining but failed to detect any skeletal malformation (data not shown). We also analyzed responses to wound healing, because mafF is expressed strongly in the skin and hair follicles (Fig. 6A), but once again, no delay of keratinization or abnormality in epithelial growth was observed in mafF homozygous mutant animals (data not shown).

DISCUSSION
In this study, we report the cloning, transcriptional, and expression analysis of the third gene encoding small Maf proteins, mafF, to clarify the activities of this family during embryonic development. As are the two other small Maf proteins, MafF is highly restricted in its temporal and tissue specificity, thus finally and definitively dispelling the early notion that the small Maf proteins were simply ubiquitously expressed partners of the more highly restricted CNC family trans-activating molecules. In fact, the converse appears to be more general, because both Nrf1 and Nrf2 are far more widely expressed than the small Maf proteins (21,22).
When we initially cloned the mouse mafF gene, structural analysis revealed that it could be transcribed from three different promoters lying close to one another but separated by about 8 kbp from the first coding exon, exon II, used by all three promoters. The mafF gene structure is thus similar to the organization of mafK, where mesodermal and neuronal transcription is differentially directed by two different promoters (34). Whether or not mafF transcription in those tissues in which it most prominently accumulates (e.g. in the lungs and skin; Figs. 5 and 6) corresponds to discrete promoter utilization is yet to be determined.
Although both mafF and mafK have now both been shown to be dispensable for viability of laboratory mice (this report; Refs. 24 and 25), that does not imply that MafF or MafK functions are meaningless. Because MafF, MafK, and MafG are each expressed in mice, chickens, and humans, one should rather surmise that in a normal competitive environment (for food tively uniformly expressed throughout the heart including the developing myocardium, and MafK is expressed primarily in the developing myocardium and valves. G, MafF and MafK are both expressed in only a subset of neurons in 12.5-dpc dorsal root ganglia, whereas MafG is uniformly and abundantly expressed there.  5. Comparison of mafF, mafG, and mafK expression patterns during embryogenesis. ␤-Galactosidase staining of mafF (left), mafG (center), or mafK (right column) homozygous mutant animals during various stages of embryonic development. A, whole mount staining of whole embryos at 6.5 dpc. The embryo is oriented to the right in each case, while extraembryonic tissue is oriented to the left; an arrow indicates the boundary between the two. B, whole mount 8.5-dpc mutant embryos. The ectoplacental cone (ec) is labeled in all three mutants. C, tissue sections of 9.5-dpc mutants. The neural tube (nt) is indicated in each embryo; all three mutants label the yolk sac endoderm (ys), but MafF appears to be most prominently expressed in the primordial gut (pg). All three small Maf proteins are expressed in the placenta (pl). D, this high magnification shows that MafF is exclusively expressed in the 9.5-dpc yolk sac endoderm (end), whereas MafG and MafK are both expressed in the endoderm as well as in red blood cells (rbc). E, this high magnification of 12.5-dpc fetal liver thin sections show that MafF is not expressed in the round (hematopoietic) cells, but that MafG and MafK both are expressed there. F, Maf F is strongly expressed in the 12.5-dpc cardiac outflow tract, whereas MafG is rela-and/or mating), MafF and MafK must confer a selective advantage to an animal bearing the gene over one in which it has been lost.
Previous studies have indicated how the small Maf proteins may be involved in both erythroid-and megakaryocyte-specific gene regulation. In consideration of the embryonic expression profile of the mafF/lacZ profile as well as the hematological analyses of mice missing the factor (Table III), it seems likely that MafF does not play a significant role in hematopoiesis. Given this conclusion then, our previous results from analysis of the mafK and mafG homozygous loss of function mutations may be interpreted as follows. In erythroid lineage cells, both MafK and MafG provide equivalent, fully redundant, activity for binding to, and activating transcription from, erythroid MARE elements. In this lineage, each homozygous mutant mouse (in mafK or mafG) displays no phenotype because of full compensation by the other small Maf protein. In contrast, MafG confers the predominant small Maf activity in megakaryocytes, and thus, MafK cannot fully compensate for MafG loss in thrombopoiesis (25). Furthermore, this explanation predicts that compound loss of MafK and MafG activity would result in both an exacerbated megakaryocytic phenotype as well as erythroid deficiencies, whereas loss of both MafF and MafG or MafF and MafK function may result in no new phenotypes. We are testing these predictions at the present time.
The results of completion of the small Maf family loss of function mutant analysis that we report here also suggests   several general conclusions about the complexity of gene regulation that might be elicited through the MARE element by this growing array of basic/zipper proteins that can interact with one another (3,9). First, MafG alone may be able to fully sustain embryonic and postnatal development, including hematopoiesis. The expression profile shown in Fig. 5 supports the notion that MafG is probably the protein fulfilling small Maf function in most tissues, and this is also most consistent with both earlier (25,35) and the present (Fig. 7) studies. Second, it is somewhat puzzling that even though MafF and MafK appear to be strongly expressed in at least a few tissues in which there is no significant accumulation of MafG (Fig. 5), there is nonetheless no phenotype associated with either individual loss of function mutation. One hypothesis that would resolve this conundrum is by proposing that a fourth small Maf protein exists, and it is this hypothetical protein that compensates for the mafF or mafK mutations in those tissues where the other two are uniquely expressed. However, numerous previous studies, including chicken and human cDNA library screens as well as screens performed here have failed to recover clones encoding any other than three known small Maf cDNAs.
If it is true that all of the small Maf proteins have been discovered, we must logically draw one of three possible conclusions. The first is that expression of the small Mafs in unique expression sites is biologically meaningless; this intellectually dissatisfying conclusion is almost certainly wrong but cannot be refuted by reference to the currently available data alone. Second, it is possible that low level expression of other small Maf proteins, below the level of detection in the lacZ staining assay, is sufficient to rescue the loss of function of a more abundant family member. Finally, it is possible that other proteins in this interactive network can partially complement the missing small Maf activity (3). With regard to this final possibility, we might imagine that c-Fos or c-Jun, which normally comprise transcription factor AP-1, are able to compensate for a loss of function mutation in mafF or mafK, because it has been shown that c-Fos or c-Jun can form heterodimers with c-Maf, Nrf1, and Nrf2 (2,7,18,19). However, heterodimers formed between AP-1 components and the large Mafs or CNC family members do not appear to exert productive regulatory effects when tested in co-transfection assays.
Finally, when members of these various leucine zipper molecules were initially cloned, it was implicitly assumed that the some or all CNC family members might be able to exert functions at MARE sites as homodimers, not requiring the cooperation of small Maf partner proteins. However, all the evidence accumulated to date suggests that homodimers of p45, Nrf1, Nrf2, Nrf3, Bach1, and Bach2 do not efficiently associate with MAREs when compared with CNC/small Maf heterodimers (11,14,16).
The complexity of all possible combinations of heterodimer formation between the small Mafs, large Mafs, CNC family, and Fos/Jun family members makes it exceedingly difficult to predict what the in vivo consequences of a mutation in any of these genes might be. This complication is exacerbated when mutations are examined in an outbred genetic background, such as the CD1 mice we used in the studies described here. To address the possible biological consequences that might arise as a result of these mutations in a more controlled environment, it would almost certainly be advantageous to examine these loss of function alleles in a constant genetic background.
To this end, we are breeding each of these small maf mutant alleles into congenic backgrounds of several different inbred strains of mice. These studies may or may not result in the presentation of new phenotypic abnormalities in the small maf mutant animals, but if each of these proteins do fulfill specific roles during development, these functions are much more likely to be revealed in a congenic genetic background.