Differential Roles of Upstream Stimulatory Factors 1 and 2 in the Transcriptional Response of Liver Genes to Glucose*

USF1 and USF2 are ubiquitous transcription factors of the basic helix-loop-helix leucine zipper family. They form homo- and heterodimers and recognize a CACGTG motif termed E box. In the liver, USF binding activity is mainly accounted for by the USF1/USF2 heterodimer, which binds in vitro the glucose/carbohydrate response elements (GlRE/ChoRE) of glucose-responsive genes. To assign a physiological role of USFsin vivo, we have undertaken the disruption of USF1 and USF2 genes in mice. We present here the generation of USF1-deficient mice. In the liver of these mice, we demonstrate that USF2 remaining dimers can compensate for glucose responsiveness, even though the level of total USF binding activity is reduced by half as compared with wild type mice. The residual USF1 binding activity was similarly reduced in the previously reported USF2 −/− mice in which an impaired glucose responsiveness was observed (Vallet, V. S., Henrion, A. A., Bucchini, D., Casado, M., Raymondjean, M., Kahn, A., and Vaulont, S. (1997) J. Biol. Chem. 272, 21944–21949). Taken together, these results clearly suggest differential transactivating efficiencies of USF1 and USF2 in promoting the glucose response. Furthermore, they support the view that USF2 is the functional transactivator of the glucose-responsive complex.

Upstream stimulatory factors (USFs) 1 belong to the basic helix-loop-helix leucine zipper (b-HLH-LZ) family of transcription factors characterized by a highly conserved C-terminal domain responsible for their dimerization and DNA binding (for review, see Ref. 1). USFs were originally described as a nuclear activity from HeLa cells able to bind and to transactivate the adenovirus major late promoter (AdML) through the canonical sequence CANNTG, referred to as an E-box motif (2). Purification of USFs revealed in fact two polypeptides of 43 (USF1) and 44 kDa (USF2), which were subsequently shown to be encoded by two distinct genes in human, rat, and mouse (3)(4)(5)(6). These two genes are ubiquitously expressed in mammalian cells. However, it is noteworthy that the relative abundance of USF1 and USF2 transcripts, as well as USF1 and USF2 proteins, varies among different cell types (7,8). Furthermore, minor forms of USF2 generated by differential splicing and alternative utilization of translational start sites have been described (8). USF1 and USF2 have been shown to bind DNA as dimers (homodimers as well as heterodimers) and to share very similar DNA binding properties (8,9). We previously reported that, in vivo, although the relative abundance between USF1 and USF2 homo-and heterodimers varies according to the tissue, the heterodimer USF1⅐USF2 seems to always represent the predominant DNA binding species (8).
Since the demonstration of USF involvement in the transcriptional activation of the AdML promoter, USFs have been reported as potential regulators of a high number of cellular genes involved in different important cellular processes (see Ref. 4). However, the respective roles of either USF1 and USF2 homodimers, or of the USF1⅐USF2 heterodimer, have usually not been well determined. Furthermore, involvement of USF is generally demonstrated by artificial ex vivo transfection experiments, which do not always reflect the actual in vivo situation. Therefore, to gain insight into the specific role of USF1 and USF2 and to evaluate the relative transcriptional activity of USF-dependent regulated genes in cells devoid of USF1, USF2, or both, we have undertaken the generation of USF1 and USF2 knock out mice. The USF2 null mice have been recently reported (10) and we describe here the generation of the USF1 gene knock out mice.
In an effort to characterize the molecular mechanisms underlying the regulation of gene expression by glucose in the liver, we previously investigated the transcriptional activation of the genes for L-type pyruvate kinase (L-PK), a glycolytic enzyme, and Spot14 (S14), a protein associated with lipogenesis (11). In the regulatory regions of these two genes, a common glucose-carbohydrate response element (termed GlRE/ChoRE) has been precisely characterized. This element consists of two E boxes separated by 5 bp. It was shown to bind in vitro a major complex containing predominantly the USF1/USF2 heterodimer (8,(12)(13)(14)(15)(16). The involvement of USFs in the glucose response was confirmed by ex vivo experiments using expression vectors (17,18) and anti-USF2 antibody microinjections (19). Finally, we were able to show that the transcriptional response of L-PK and S14 genes was altered in USF2 Ϫ/Ϫ mice (10). In this paper, we present the transcriptional response of L-PK and S14 gene to glucose in the USF1 Ϫ/Ϫ mice and discuss the results considering the previously reported data obtained in the USF2 Ϫ/Ϫ mice.

EXPERIMENTAL PROCEDURES
Gene Targeting and Generation of USF1 Ϫ/Ϫ Mice-The USF1 genomic fragments were all generated from clones isolated from an ES-129 embryonic stem cell line murine genomic library (4). ES culture and embryo manipulation were performed as described (20).
A 4-kb SpeI-SpeI fragment, encompassing exons 2, 3, and part of exon 4, was blunted, modified by the addition of NotI linkers, and subcloned into the 5Ј NotI site of the reported IRES-␤geo-containing vector subcloned into SK (21). This plasmid was SalI digested and blunted, and the blunted 3Ј-USF1 homologous fragment, consisting of a 2.7-kb XhoI-XbaI fragment, was inserted. This fragment contains part of exon 10 and 3Ј flanking sequences of the USF1 gene, leading thus to a 3-kb deletion of genomic sequences between exon 5 and exon 10. The resulting targeting construct was linearized at the unique XhoI site, and 60 g was used to electroporate 1 ϫ 10 7 D3-ES cells (22), which were cultured on mitomycin-treated embryonic fibroblast feeder layers. DNA from clones surviving G418 selection (200 g/ml) were individually analyzed by Southern blot. For 5Ј and internal analyses, DNA was EcoRI-digested and hybridized either with a 5Ј external probe consisting of a genomic 0.7-kb EcoRI-NaeI fragment (see Fig. 1A) or with an internal probe consisting of the 2.7-kb XhoI-XbaI 3Ј homologous fragment. For 3Ј analysis, DNA was BamHI digested and hybridized with a 3Ј external probe consisting of a genomic 2.6-kb EcoRI-EcoRI fragment spanning downstream sequences of the USF1 gene. External and internal probes were further used to ensure the absence of additional random integration of the targeting construct (data not shown).
ES cells from two positive targeted clones (76 and 83) were injected into blastocyste derived from C57BL/6J mice. Chimeric males derived from both clones were mated to wild type C57BL/6J females for the germ line transmission. Mice heterozygous for the gene-targeting event were then used to generate homozygous mutant USF1 Ϫ/Ϫ mice. As the phenotype and the metabolic results were identical with animals derived from both ES cell clones, the experiments were then performed only with the F2 descendant animals from the clone 83. To generate the double heterozygote USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice, USF1 ϩ/Ϫ mice were bred with the previously reported USF2 ϩ/Ϫ mice (10). Genotyping of offspring was performed by Southern blot analysis with 10 g of EcoRIdigested mouse tail DNA.
Nutritional Treatment and Metabolic Parameters of the Animals-For metabolic studies, animals were fed a high carbohydrate diet for 18 h after a 24-h fasting. Mice were sacrificed between 10 and 12 a.m., and tissue samples were stored at Ϫ80°C. Blood samples were collected from the orbital sinus.
Glucose level was determined with glucose oxidase using the Sigma kit, and the level of glucox 6-P in crude liver extract was measured as described previously (10).
RNA Analysis-RNAs were purified by a modified guanidium chloride procedure from frozen liver, and Northern blot analysis was conducted as described before (23). For detection of L-PK mRNAs, a rat cDNA clone (clone G4) encompassing the coding sequence was used (24). For S14 detection, a polymerase chain reaction-made fragment corresponding to the murine coding sequence was used as a probe (25).
Each Northern blot was stripped and reprobed with a ribosomal 18 S cDNA to check for the integrity and the amount of loaded RNAs. The amount of specific mRNA was quantified using a PhosphorImager (Molecular Dynamics).
For electrophoretic mobility shift assays (EMSA), the DNA binding reaction was performed as described previously (8) in the presence of 30 g of whole cell extract proteins, 10 g of poly(dI-dC), and 0.1-0.5 ng of end-labeled MLP oligonucleotide. For supershift experiments, USF1 and USF2 antibodies were included in the binding reactions (8). Quality and quantity of the extracts were monitored by analyzing the ubiquitous NFY binding activities. DNA binding complexes separated on nondenaturing polyacrylamide gels were quantified using a Phospho-rImager (Molecular Dynamics).
Data Analysis-Statistical analysis was performed by the Student's t test for unpaired data using the StatView software. The significance has been considered at p Ͻ 0.05.

Generation and Characterization of Homozygous USF1-deficient
Mice-To inactivate the USF1 gene, a replacement vector was constructed with 6.7 kb of ES-129-derived genomic fragment and a selection cassette disrupting exon 4 (Fig. 1A). Ensured that the USF1 gene was active in ES cells, we carried out a promoter trap strategy to maximize the targeting frequency. To this end, the promoterless IRES-␤geo cassette was chosen as the selection cassette. This cassette allows, via an IRES element, for efficient translation of the ␤geo fusion transcript. It leads to the synthesis of a fusion protein with both ␤-galactosidase and neomycin activities (21). After electroporation and selection of recombinant ES cells, homologous recombinants were detected by Southern analysis as described under "Experimental Procedures." Among 155 colonies screened, 76% were appropriately targeted. This result confirms previous data showing that the use of IRES-containing constructs may allow homologous recombination events to be isolated at a very high frequency (21). After transmission of the mutation, heterozygous FIG. 1. Inactivation of the USF1 ؊/؊ gene. A, schematic representation (not to scale) of the genomic structure of the mouse USF1 gene (top), the targeting vector (middle), and the targeted gene (bottom). In the latter one, the two ATGs leading to the dicistronic transcript are indicated. The ten exons are represented as boxes, with shaded boxes corresponding to the b-HLH-LZ domain of the USF1 protein. EcoRI restriction sites are noted (E), and expected fragment sizes of the wild type and mutant allele after EcoRI digestion and hybridization with a 5Ј external probe (solid bar) are indicated. B, Southern blot analysis of offspring derived from heterozygote intercrosses. Tail DNA was digested by EcoRI and hybridized with the 5Ј external probe. The USF1 genotype is indicated above each lane. C, Western blot analysis of USF1 in fetal brain cellular extracts from ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ animals.
progeny, which was normal and fertile, was intercrossed to produce homozygous offspring. The presence of the mutated gene in the F2 offspring was routinely checked by Southern analysis of tail DNA. As shown in Fig. 1B, hybridization of EcoRI-digested genomic DNAs with the indicated 5Ј external probe lead to wild type alleles of 10.8 kb and mutated alleles of 5.1 kb.
Disruption of the USF1 gene resulted in the complete absence of USF1 protein in USF1 Ϫ/Ϫ animals. This was ascertained by immunoblotting of cellular proteins of USF1 Ϫ/Ϫ, USF1 ϩ/Ϫ, and USF1 ϩ/ϩ animals with an anti-USF1 specific antibody (Fig. 1C). The 43-kDa USF1 protein was detected only in extracts from heterozygous and wild type animals. The reduced level of USF1 in USF1 ϩ/Ϫ animals, as compared with the ϩ/ϩ animals, suggested the failure of the wild type USF1 allele to compensate for the mutation.
Characterization of Homozygous USF1 Ϫ/Ϫ Knock Out Mice-Out of the first 239 F2 offspring genotyped, 19.7% were homozygous (Ϫ/Ϫ) for the disrupted allele, and 24.7% were wild type (ϩ/ϩ), which was compatible with the expected Mendelian frequency. At birth, the homozygotes seemed to be indistinguishable from wild type and heterozygous littermates. Furthermore, in contrast to the severe phenotype of the USF2 Ϫ/Ϫ mice (10), USF1 Ϫ/Ϫ mice developed and reproduced without any obvious alteration. No behavioral defects or anatomical abnormalities in any major organs were seen.
USF Binding Activity Is Accounted for by USF2 Homodimers in USF1-deficient Mice-To determine the changes in the composition of USF binding activity, liver cellular extracts were used in EMSAs with a double-stranded oligonucleotide containing the canonical E box from the AdML promoter ( Fig. 2A). As previously reported (10), the major USF binding activity in the ϩ/ϩ mice liver was predominantly accounted for by USF1/ USF2 heterodimer. USF1 and USF2 homodimers, visualized as weak bands after supershift with either anti-USF2 or USF1 antibody, respectively, seemed to be minor complexes. In contrast, in USF1 Ϫ/Ϫ cellular extracts, USF binding activity, visualized as a fainter and slightly upper migrating band, was only accounted for by the USF2 homodimer. Indeed, USF2 antibody fully displaced this complex, whereas the USF1 antibody had no effect. Quantification of this residual USF binding activity was performed by EMSA with an internal control consisting of the binding site for the ubiquitously expressed NFY factor (Fig. 2B). This analysis revealed that the level of USF2 binding activity was reduced to 42% (n ϭ 6, p ϭ 0.0001) to that of the wild type. This result demonstrates that, in vitro, the only detectable binding activity on the E boxes, in cell extracts devoid of USF1, is USF2 and that probably USF2 gene expression is not up-regulated in the liver of USF1 KO mice. This result was further confirmed by Western blot analysis showing no change in the amount of USF2 in liver cellular extracts of either Ϫ/Ϫ or ϩ/ϩ mice (Fig. 2C).

Hepatic Content of L-PK and Spot14 mRNAs Is Not Modified in USF1-deficient Mice as Compared with Wild Type Mice-To
determine the impact of USF1 deficiency on glucose responsiveness of hepatic gene expression, a series of metabolic analyses was performed on wild type and USF1 Ϫ/Ϫ mice. After a 24-h fasting period, animals were refed with a high carbohydrate diet for 18 h. Glucose homeostasis was assessed by measuring blood glucose level, and hepatic glucose utilization was monitored by quantitating the level of glucose-6P. Both values were identical in normal and mutated mice (glycemia of 1.43 Ϯ 0.3 (n ϭ 9) and 1.41 Ϯ 0.23 (n ϭ 11) g/liter glucose, for the ϩ/ϩ and Ϫ/Ϫ groups, respectively; and glucose 6-P concentration of 8.2 Ϯ 2.0 (n ϭ 9) and 8.3 Ϯ 2.6 (n ϭ 11) nmol/mg of proteins for the ϩ/ϩ and Ϫ/Ϫ groups, respectively). This indicated an apparent normal glucose uptake and utilization by the liver of USF1 Ϫ/Ϫ mice. Total liver RNAs were then analyzed by Northern blot to assess the content of L-PK and S14 mRNAs (Fig. 3A). Quantification of this analysis (Fig. 3B) revealed that the amount of L-PK and S14 mRNAs in the liver of the USF1 Ϫ/Ϫ mice was not statistically different from that of the ϩ/ϩ mice.

Hepatic Content of L-PK and Spot14 mRNAs Is Not Altered in Double Heterozygote Mice, Whereas USF Binding Activity Is
Reduced 2-Fold-The absence of hepatic alteration of L-PK and S14 gene transcription in response to a hyperglucidic diet in USF1 Ϫ/Ϫ mice suggested that the remaining USF2 homodimers in the liver cells were able to substitute for the USF1/USF2 heterodimer. Interestingly, this substitution took place while the amount of USF binding activity was reduced to 42% of the total USF binding activity. We therefore asked whether this observation was specific to USF2 or if reducing the quantity of the heterodimer would have also led to unchanged L-PK and S14 mRNA levels. To address this question, we submitted wild type and double heterozygote USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice to the above described feeding protocol. We first compared by EMSA the amount of USF binding activity in liver cellular extracts of wild type and USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice. Fig.  4A shows a representative EMSA experiment along with the quantification of USF binding activity. This result established that the USF binding activity, which was mainly accounted for by the heterodimer, was reduced to 46% in the double heterozygote mice. We then measured, by Northern blot analysis, the amount FIG. 2. Analysis of USF binding activity and USF content in liver cellular extracts from ؉/؉ and ؊/؊ animals refed a carbohydrate-rich diet for 18 h. A, EMSA was performed using the radiolabeled MLP oligonucleotide in the presence of liver cellular extracts from ϩ/ϩ and Ϫ/Ϫ mice. Selective depletion of the complexes was realized by adding preimmune serum (Ϫ) or specific anti-USF antibodies (Ab) in the binding reactions. The asterisk indicates nonspecific protein complexes. B, EMSA was performed as above with an internal NFY binding site in the binding reaction. C, Western blot analysis of liver cellular extracts with USF2 antibody in USF1 Ϫ/Ϫ mice. of L-PK and S14 mRNAs in the liver of USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice as compared with wild type mice. As presented in Fig. 4B, the amount of L-PK and S14 mRNAs was comparable in the liver of wild type and USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice. Thus, the response of the L-PK and S14 genes was not altered in USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice despite the reduced USF binding activity. DISCUSSION In the USF1 Ϫ/Ϫ mice presented in this paper and in the previously reported USF2 Ϫ/Ϫ mice (10), we have investigated the regulation of two glucose-responsive genes, namely the L-type pyruvate kinase and Spot 14. Both genes were shown to contain in their regulatory region a common sequence responsible for mediating the glucose response. This sequence consists of two 5Ј CACGTG type E box motifs separated by 5 bp. It is able to bind in vitro members of the USF family (8,(12)(13)(14)(15)(16). Several lines of evidence from previous ex vivo and in vivo studies support the importance of this cis-acting sequence in promoting the glucose responsiveness. It was of prime interest to test the involvement of the trans-acting proteins described as binding this sequence in vitro. In the liver of wild type mice, we previously reported that the USF binding activity is mainly accounted for by the USF1/USF2 heterodimer (10). By EMSA studies, we demonstrated, here and in the previously reported USF2 Ϫ/Ϫ KO mice (10), that the hepatic USF binding activity is accounted for by the remaining homodimer species, i.e. USF1 homodimer in USF2 KO mice and USF2 homodimer in USF1 KO mice. In addition, the amount of this homodimer in the liver, as assayed by gel shift and Western blot experiments, is consistent with an unchanged expression of the remaining gene. In other words, neither liver USF1 gene expression in USF2 KO mice nor USF2 in USF1 KO mice seems to be overexpressed consequently to the mutation.
These results in adult liver are different from those recently reported by Sirito et al. (26) in fetal fibroblasts from USF2 Ϫ/Ϫ and USF1 Ϫ/Ϫ mice. The authors observed that the USF2 Ϫ/Ϫ mice contain a reduced level of USF1, whereas the USF1 Ϫ/Ϫ mice seem to compensate USF1 deficiency by increased USF2 synthesis. It seems therefore that the asymmetrical cross-regulation between USF genes during fetal life is replaced after birth, at least in the liver, by an independent pattern of USF1 and USF2 gene expression.
We previously demonstrated that in USF2 Ϫ/Ϫ mice refed a hyperglucidic diet for 18 h, L-PK and S14 gene expression was dramatically reduced (by 3.8-and 6.8-fold, respectively). Therefore, the glucose responsiveness of glucose-responsive genes in the liver, impaired in the presence of USF1 homodimers in FIG. 3. L-PK and S14 mRNAs content in liver of ؉/؉ and ؊/؊ animals. A, Northern blot analysis in animals fasted for 24 h and refed a carbohydraterich diet for 18 h. B, quantitated results are expressed as means Ϯ SD of nine mice for both groups. The values obtained in USF1 Ϫ/Ϫ mice are not statistically different from those in USF1 ϩ/ϩ mice (p ϭ 0.07 for L-PK and p ϭ 0.94 for S14).
FIG. 4. USF DNA binding activity and mRNA content for L-PK and S14 in the liver of wild type and USF1 ؉/؊ USF2 ؉/؊ mice. A, EMSA was performed using the radiolabeled MLP oligonucleotide in the presence of liver cellular extracts from wild type and USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice. Quantitated results are expressed as means Ϯ SD of six mice for both groups (p ϭ 0.0001). B, Northern blot analysis of liver RNA from wild type and USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice fasted for 24 h and then refed a carbohydrate-rich diet for 18 h. Quantitated results are expressed as means Ϯ SD of seven ϩ/ϩ mice and nine USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice. The values obtained in USF1 ϩ/Ϫ USF2 ϩ/Ϫ mice are not statistically different from those in USF1 ϩ/ϩ (p ϭ 0.64 for L-PK and p ϭ 0.18 for S14).
USF2 Ϫ/Ϫ mice, is normal in the presence of USF2 homodimers in USF1 Ϫ/Ϫ mice. Because the glucose response of the L-PK and S14 genes is also normal in double heterozygous mutants, we can conclude that 40 -50% of residual USF binding activity is sufficient to sustain a normal glucose response when consisting of either USF2 homodimers or USF1/USF2 heterodimers but not of USF1 homodimers. In other words, USF1 seems to be intrinsically less efficacious than USF2, at least as a transactivating factor of the glucose-response complex. The weak transcriptional effect of USF1 is reminiscent of the results reported by Desbarats et al. (27). These authors demonstrate that the activity of USF1 on isolated E boxes in vivo is often poor and that the distance between E boxes and start sites of transcription can be a critical factor. This observation could explain why response of the S14 gene to glucose is much more affected than that of the L-PK gene in USF2 Ϫ/Ϫ mice. Indeed, the E boxes are located far upstream from the start site of transcription (Ϫ1422 bp) in the S14 gene and in a very proximal position (Ϫ144 bp) in the L-PK gene. Furthermore, Carter et al. (28) recently called into question the traditional view of USF1 homodimer as a transcriptional activator.
In an effort to determine the selective utilization of b-HLH-LZ proteins at the Ig heavy chain enhancer, the authors surprisingly observed that USF1 had a negative effect on the activity of the IgH. They hypothesize that repression may be a consequence of the absence of a strong activation domain in USF1. Different studies suggest that USF1 could act as a repressor by competing in vivo with positive-transacting factors (29,30).
The transactivating activity of USF1 and USF2 dimers could be intrinsically different and/or it could be brought out by differential interaction of USF1 and USF2 with other regulatory proteins. The highly divergent USF1 and USF2 NH 2terminal domains, encompassing the transcriptional activation domain, could specify these interactions. Alternatively, a role for the b-HLH-LZ domain could be evoked. In this respect, it is interesting to note that a specific cross-talk between USFs and other transcription factors begins to be documented. Blanar and Rutter (31) have first reported the in vitro interaction of USF2 with c-Fos. More recently, by different cloning strategies, specific interactions have been reported, on one hand, between USF1 and the b-Zip Fra1 factor (32), the b-HLH-E47 protein (33), and the winged HTH Ets-1 factor (34); on the other hand, specific interactions have been reported between USF2 and the c-Maf1 factor (35). So far, however, further experiments are needed to assign a physiologic role to these physical interactions. Finally, Roy et al. (36) recently reported the molecular cloning of the TFII-I transcriptional factor that physically and functionally interacts with USF1.
It will be interesting to determine whether the differential transactivating properties of USF1 and USF2 dimers can occur on other USF-dependent genes and, in this case, to what extent they could target different sets of responsive genes in different cell types and in response to various stimuli.
Although the results discussed before suggest that the USF1 homodimers are not required for normal embryonic development, postnatal growth, and transcriptional response to glucose and that they cannot by themselves totally compensate for USF2 deficiency, their role is nevertheless attested by embryonic lethality of double knock out mice deficient in both USF1 and USF2 (26). 2 Preliminary data also indicate that the phenotype is more severe in animals with only a residual USF1 intact allele (USF1 ϩ/Ϫ USF2 Ϫ/Ϫ), which die earlier during in utero development, than in those with a residual USF2 allele (USF1 Ϫ/Ϫ USF2 ϩ/Ϫ), which generally die immediately before or after birth. 2 This observation is consistent with a better efficacy of USF2 than of USF1. However, the underlying mechanism is here uncertain because, as mentioned above, Sirito et al. (26) have shown that USF1 concentration is decreased in USF2-deficient mice, whereas USF2 is increased in USF1-deficient mice.
In conclusion, the USF knock out mouse models have allowed us to demonstrate a differential role of USF1 and USF2 in the transcriptional response of liver genes to glucose. Indeed, a normal glucose responsiveness seems to require the presence of either USF1/USF2 heterodimers or USF2 homodimers, even in mice with total USF binding activity reduced by half, whereas USF1 homodimers give rise to a delayed response. This result could reflect an intrinsic difference in the transactivating efficiency of USF isoforms. Accordingly, consequences of USF2 deficiency on embryonic development are more severe than those of USF1 deficiency. Further investigations will determine how transactivation of glucose responsiveness by USF2containing dimers is finely tuned by glucose.