(Received for publication, February 26, 1997, and in revised form, May 14, 1997)
From the Institut Cochin de Génétique
Moléculaire, U.129 INSERM Unité de Recherches
en Physiologie et Pathologie Génétiques et
Moléculaires and ** U.257 INSERM Laboratoire de
Génétique Cellulaire et Moléculaire, Université
René Descartes, 24 rue du Faubourg Saint-Jacques,
75014 Paris, France
Upstream stimulatory factors (USF) 1 and 2 belong
to the Myc family of transcription factors characterized by a
basic/helix loop helix/leucine zipper domain responsible for
dimerization and DNA binding. These ubiquitous factors form homo- and
heterodimers and recognize in vitro a CACGTG core sequence
termed E box. Through binding to E boxes of target genes, USF factors
have been demonstrated to activate gene transcription and to enhance
expression of some genes in response to various stimuli. In particular,
in the liver USF1 and USF2 have been shown to bind in vitro
glucose/carbohydrate response elements of glycolytic and lipogenic
genes and have been proposed, from ex vivo experiments, to
be involved in their transcriptional activation by glucose. However,
the direct involvement of these factors in gene expression and nutrient
gene regulation in vivo has not yet been demonstrated.
Therefore, to gain insight into the specific role of USF1 and USF2
in vivo, and in particular to determine whether the USF
products are required for the response of genes to glucose, we have
created, by homologous recombination, USF2 /
mice. In this paper,
we provide the first evidence that USF2 proteins are required in
vivo for a normal transcriptional response of L-type pyruvate
kinase and Spot 14 genes to glucose in the liver.
Glucose is a major source of metabolic energy for all mammalian cells. In the liver, glucose metabolism begins or ends with the movement of glucose into or out of the hepatocyte. When hormonal conditions favor glucose utilization (high insulin, low glucagon), glucose entry is facilitated along a downhill gradient through the specific GLUT2 transporter, and intracellular glucose is rapidly phosphorylated (1, 2).
In response to the glucose signal, transcription of several genes implicated in glucose metabolism is activated (for review see Refs. 3 and 4). Two such glucose-responsive genes have been extensively studied in the liver, i.e. the genes for L-type pyruvate kinase (L-PK),1 a glycolytic enzyme, and Spot 14 (S14), a protein associated with lipogenesis. Transient transfection assays in primary hepatocytes have allowed for the identification of a common glucose response element, termed GlRE by our group and ChoRE, for carbohydrate response element, by the group of H. Towle, in the regulatory regions of the L-PK and S14 genes (5-8). The functional importance of the GlRE has been further demonstrated in vivo using transgenic mice (9).
The L-PK GlRE (located between nucleotides 168 and
144, with
respect to the liver-specific start site of transcription) and the S14
ChoRE (between nucleotides
1448 and
1422) present a similar
arrangement consisting of two E boxes separated by 5 base pairs (10).
The so-called E box consensus CACGTG is the recognition site for
members of the Myc family of transcription factors characterized by a
highly conserved C-terminal basic/helix-loop-helix/leucine-zipper domain responsible for dimerization and DNA binding (11). Through binding to their DNA target as homo- or heterodimers, these factors are
known to regulate a variety of genes (12, 13). In the liver, the
predominant members of this family able to recognize the L-PK GlRE
in vitro are USF factors (14). These USF factors were
originally described as a binding activity able to interact and to
transactivate the adenovirus major late promoter (15). Purification of
this activity revealed the presence of two polypeptides of 43 (USF1)
and 44 kDa (USF2a) (16). These two forms of USF are encoded by distinct
genes that have been cloned and characterized in the mouse (12, 13, 17,
18). These genes are widely expressed in mammalian tissues, and we have
assessed that, in the liver, the USF1/USF2a heterodimer accounted for
more than 65% of the USF binding activity (14). Besides, we have shown that, when transfected into hepatoma cells, USF isoforms were able to
interfere with the glucose response of the L-PK gene (19).
To test for the possible involvement of USF2 in the glucose-mediated
induction of hepatic gene expression in vivo and to fully reveal its biological role, we sought to examine mice with a null mutation for USF2. To this end, we have created a mouse USF2 /
line
by disrupting the USF2 locus by gene targeting. In this paper, we
present the first model of USF2
/
mice and provide evidence that,
in the liver, USF2 proteins are needed in vivo for a normal transcriptional response of the L-PK and S14 genes to glucose.
The USF2
genomic fragments were all generated from clones isolated from a
murine ES-129 embryonic stem cell line genomic library (12). ES culture
and embryo manipulation were performed as described (20).
A 7.5-kb XhoI-NotI fragment, encompassing part of
exon 7 and the almost totality of intron 7, was blunted, modified by
the addition of NotI linkers, and subcloned into the 3
NotI site of the previously reported internal ribosomal
entry site-
geo containing vector (21). This plasmid was
SalI-digested, and the 5
USF2 homologous fragment,
consisting of a 1.8-kb XhoI-XhoI fragment
spanning from intron 1 to exon 7, was inserted. The resulting targeting
construct was linearized at the unique XhoI site, and 60 µg was used to electroporate 1 × 107 CCB-ES cells
derived from 129Sv/J mice (kindly provided by Martin Evans, Cambridge,
UK), which were cultured on mitomycin-treated embryonic fibroblast
feeder layers. DNA from clones surviving G418 selection (200 µg/ml)
were individually analyzed on Southern blot. DNA was either
EcoRI-digested and hybridized with a 5
external probe
consisting of a genomic EcoRV-BamHI fragment in
the 5
regulatory sequences of the USF2 gene (see Fig. 1A)
or EcoRV-digested and hybridized with a 3
external probe
consisting of a polymerase chain reaction-amplified fragment
encompassing exon 8 to exon 10. ES cells from positive clones were
injected into blastocysts derived from C57BL/6J mice (20). Chimeric
males, as judged by the agouti coat color, were mated to wild-type
C57BL/6J female for the germ line transmission. Mice heterozygous for
the gene targeting event were then used to generate homozygous mutant
USF2
/
mice. Genotyping of offspring was performed by Southern blot analysis with 10 µg of EcoRI-digested mouse tail DNA.
Nutritional Treatment and Metabolic Parameters of the Animals
For metabolic studies, animals were fed a high
carbohydrate diet, for either 18 h (after a 24-h fasting), 40 h, or 5 days. Mice were sacrificed between 10 and 12 a.m. and
tissue samples were stored at 80 °C. Each animal was referenced
with a double number, the first one indicating the litter and the
second one the animal figure.
Serum insulin levels were determined using an insulin radioimmunoassay kit (Behring Diagnostics) with human insulin as standard, and glucose levels were determined with glucose oxidase using a Sigma kit. The level of glucose 6-P in crude liver extract was measured as described previously (22). Liver samples were homogenized in 1 ml of 6% (w/v) perchloric acid. Denatured proteins were sedimented by centrifugation for 10 min at 17,000 × g, and the supernatant was neutralized with KOH in the presence of 20 µl of phenol red. The resulting potassium perchlorate was discarded by centrifugation (10 min at 12,000 × g). Glucose 6-P in 0.45 ml of the neutralized extract was quantified by measuring the change in absorbance at 340 nm in an assay system of 1 ml containing 1 mM NADP+, 50 mM Tris-HCl, pH 7.3, and 1.7 unit of glucose-6-P dehydrogenase (Torula Yeast). A glucose 6-P control was used to assess the reliability of the assay. Subsequently, the amount of total proteins for each liver sample was determined using the Coomassie Blue staining, and the results were expressed in nmol of glucose 6-P per mg of proteins.
Oral Glucose Tolerance TestAfter an overnight fasting,
/
and +/+ animals were anesthetized with an intraperitoneal
injection of avertin. A 40-mg glucose load was administered into the
stomach through a cannula. Blood was sampled from the tail vein at time
0, 15, 30, 45, 60, and 75 min after glucose ingestion, and glucose was
determined using a Glucometer II and Glucostix strips (Ames).
RNAs were purified by a modified guanidium chloride procedure from frozen liver, and Northern blot analysis was conducted as described previously (23). For detection of L-PK mRNAs, a rat cDNA clone (clone G4), encompassing the coding sequence was used (24). For GLUT2 detection, a rat GLUT2 cDNA was kindly provided by Bernard Thorens (Lausanne, Switzerland). 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.
Protein AnalysisFetal liver samples assayed for L-PK content were homogenized in 10 mM Tris, pH 7.4, 1 mM phenylmethylsulfonyl fluoride, 0.1% aprotinin, 1 µg/ml leupeptin, and 25 mM benzamidine. After centrifugation (15,000 × g, 1 h, 4 °C), the supernatant proteins were quantified using the Coomassie Blue stain with bovine serum albumin as standard. Proteins were subjected to electrophoresis in a 12% (w/v) SDS-polyacrylamide gel, and the separated proteins were electrotransferred to nitrocellulose membranes in the presence of 10% (v/v) methanol. Efficiency of transfer was checked with Ponceau S staining. After blocking nonspecific protein binding sites for 1 h at room temperature in Tris-buffered saline-Tween reagent (TBS-T) containing 5% (w/v) non-fat dry milk, the blot was incubated with L-PK antibody (26) at a 1:1000 dilution for 1.5 h at room temperature. Following three 15-min washes in TBST, the secondary antibody, peroxidase-conjugated swine anti-rabbit (Dako), was added at a 1:1000 dilution for 1 h at room temperature. The blot was washed again as above, and peroxidase activity was detected by fluorography using the enhanced chemiluminescence system (Amersham Corp.). To ensure comparable loading of the samples, the blot was incubated with an anti-annexin V antibody. Specific bands were quantified by scanning the autoradiographs with a Shimadzu densitometer.
Adult liver samples assayed for USF2 content were homogenized in 20 mM Hepes, pH 8, 1.5 mM MgCl2, 0.2 mM EDTA, 1 mM dithiothreitol, 25% (v/v) glycerol, 0.2% (v/v) Nonidet P-40, and protease inhibitor mixture tablets "COMPLETE"TM (Boehringer Mannheim).
Western blot analyses were performed as described above, using affinity purified USF2 1-49 antibodies at a 1:200 dilution (14).
For electrophoretic mobility shift assays, the DNA binding reaction was performed as described previously (14) in the presence of 20 µg of whole cell extract proteins, 10 µg of poly(dI-dC), and 0.1-0.5 ng of end-labeled major late promoter oligonucleotides. For supershift experiments, USF1 and USF2 antibodies were included in the binding reactions (14). Quality of the extracts was checked by analyzing the ubiquitous NFY binding activities. DNA-binding complexes separated on nondenaturing polyacrylamide gels were quantified using a PhosphorImager (Molecular Dynamics), and the relative amount of radioactivity was measured by volume integration with the ImageQuant software.
To inactivate the USF2
gene, a replacement vector was constructed with 9.3 kb of ES-129
derived genomic fragment and a selection cassette disrupting exon 7 (Fig. 1A). Ensured that the
USF2 gene was active in embryonic stem (ES) cells, we carried out a
promoter trap strategy to maximize the targeting frequency. To this
end, the promoterless internal ribosomal entry site-geo cassette was chosen as the selection cassette. This cassette allows, via an internal
ribosomal entry site element, for efficient translation of the
geo
fusion transcript which leads to the synthesis of a fusion protein with
both
-galactosidase and neomycin activities (21). Successful
targeting should abolish the transcriptional function of USF2 by
disrupting the basic domain, encoded by exon 7, responsible for its DNA
binding. After electroporation and selection of recombinant ES cells,
homologous recombinants were detected by Southern analysis. Among them,
one targeted stem cell clone contributed particularly to the germ line
of chimeric mice. After transmission of the mutation, heterozygous
progeny, which were normal and fertile, were 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 5
external probe led to wild-type
alleles of 12.5 kb and mutated alleles of 8.5 kb.
Of
the 110 F2 offspring analyzed after cesarean section at 19 days
post-coitum, 21.8% were homozygous (/
) for the disrupted allele
and 25.4% were wild type (+/+), indicating that USF2 deficiency did
not affect embryonic development. However, after birth, survival of
homozygous mice was markedly affected by disruption of the USF2 gene.
Indeed, only 20% homozygous mice survived a few hours after birth. The
USF2
/
mice were significantly smaller in size than their
littermates. This decrease in body weight (approximately 20%) was
already evident at the embryo stage of 17.5 days and either stabilized
or dramatically increased after birth. Based on this body size
criterion, we were able to significantly improve survival of
/
mice, by reducing the litter size. Still, the lifespan of surviving
mice was variable. This increase in the survival of USF2
/
mice
suggested that part of the lethal phenotype could be due, at least, to
competition for food as homozygotes seemed more apathetic than their
littermates. Phenotype of the knock-out mice will be presented in
detail elsewhere.
Disruption of
the USF2 gene resulted in the complete absence of USF2 protein in /
animals. This was ascertained by immunoblotting of liver nuclear
proteins of
/
,
/+, and +/+ animals with an anti-USF2 specific
antibody (Fig. 2A). The 44-kDa
USF2a protein was detected only in extracts from heterozygous and
wild-type animals. The reduced level of USF2 in
/+ animals, as
compared with the +/+ animals, suggested the failure of the wild-type
USF2 allele to compensate for the mutation. To determine the changes in
the composition of USF binding activity, the same nuclear extracts were
used in electrophoretic mobility shift assays with a double-stranded oligonucleotide containing the canonical E box from the adenovirus major late promoter.
As visualized in Fig. 2B, in +/+ extracts, a major band
migrating as the control USF1/USF2 heterodimer of the rat liver nuclear extract was observed. The same complex, although less intense, was
detected in the /+ extracts. A fainter and slightly lower band was
observed in the
/
extracts which was reminiscent to that observed
in rat liver USF2-immunodepleted extracts (14). To precisely determine
the nature of this complex, we used USF1 and USF2 antibodies (Fig.
2C). As previously observed, USF1 and USF2 homodimers,
visualized as weak bands after supershift with either USF1 or USF2
antibody, appeared to be minor complexes in the liver nuclear extracts
of the +/+ and
/+ mice. On the contrary, in
/
nuclear extracts,
the use of USF1 antibody fully displaced the observed complex, and the
USF2 antibody had no effect. This revealed that, in vitro,
the only detectable binding activity on the E boxes, in extracts from
animals devoid of USF2, was the USF1 homodimer. By assessing the
quantity of USF1 homodimer in
/
extracts, we could speculate that
the expression of USF1 gene was poorly affected by the absence of USF2.
To test this, RNA samples from the liver of
/
and +/+ mice were
assayed by Northern blot using specific USF1 probe. This analysis
revealed indeed no change in the USF1 gene expression (data not shown).
This result was consistent with our previous observation indicating the
absence of E box in the USF1 gene sequence (13).
To determine the
impact of USF2 deficiency on L-PK gene expression, our first
experiments were conducted on fetuses of several littermates. After
cesarean section at the end of gestation, each fetal liver was
homogenized, and the content of L-PK protein was assessed by Western
blot analyses using specific L-PK antibody. The signals were
quantitated and averaged out into /
,
/+, and +/+ groups after
genotyping. As shown on the histogram of Fig. 3, the amount of L-PK protein was
significantly reduced in liver of
/
fetuses.
To study further the glucose responsiveness of hepatic gene expression,
a series of metabolic analyses was performed on wild-type and surviving
USF2 /
mice. After an 18-h refeeding period with a high
carbohydrate diet, animals were sacrificed, and total liver RNAs were
prepared and analyzed by Northern blot to assess the content of L-PK
and S14 messenger RNAs. This analysis revealed that, in the liver of
all the
/
animals tested, the amount of both L-PK and S14 mRNAs
was markedly lowered as compared with the +/+ animals (Fig.
4A). In ad libitum
fed animals, the amount of both L-PK and S14 mRNAs was not
significantly different (data not shown).
After a 40-h feeding period with a high carbohydrate diet, the
difference in the level of L-PK and S14 mRNAs was still obvious except for the 17.1 mouse that did contain relatively high level of
mRNAs (Fig. 4B). However, after 5 days on this high
carbohydrate diet, /
and +/+ animals displayed similar levels of
L-PK and S14 mRNAs (Fig. 4C).
To assign the diminution in the level of L-PK and S14 mRNAs, observed 18 h after glucose refeeding, to the transcriptional state of the genes rather than to a putative alteration of glucose metabolism, several steps, from blood glucose disposal to cellular glucose utilization, were carefully checked.
Blood glucose disposal was first assessed by performing an oral glucose
tolerance test in fasted mice. As shown in Fig.
5, after the glucose load, glucose
accumulation was quite similar in +/+ and /
animals. To further
challenge glucose homeostasis in
/
animals, blood glucose
concentration was assayed in fasted and refed animals. The absence of
USF2 seemed to have no effect on the glycemia of either fasted animals
(series not shown) or refed animals (1.8 ± 0.5 (n = 7) and 1.6 ± 0.4 (n = 10) g/liter glucose for
the
/
and +/+ groups, respectively, Table
I). The next step was to confirm that
insulin was correctly synthesized and secreted in response to oral
glucose, as suggested from the normal plasma glucose levels. As
monitored, neither the amount of pancreatic insulin messenger RNA (not
shown) nor the plasma insulinemia (Table I) was modified in USF2
/
refed mice.
|
The second part was to evaluate glucose entry and phosphorylation,
which in hepatocytes involve the specific GLUT2 transporter, and
hexokinase IV or glucokinase (1, 2). This coupled
transport/phosphorylation system forms part of a glucose sensing
apparatus that responds to subtle changes in blood concentration.
Therefore, GLUT2 gene expression was assessed by Northern blot analysis
and, in fact, as shown in Fig. 6, the
hepatic amount of GLUT2 mRNA remained unchanged in USF2 /
mice.
To monitor glucose phosphorylation into glucose 6-P, we measured
directly the intracellular concentration of glucose 6-P in the liver of
the animals. The results, presented in Table I, demonstrated that the
glucose 6-P concentration was roughly similar in both
/
and +/+
groups, 6.2 ± 2.0 (n = 10) and 5.9 ± 1.6 (n = 7) nmol/mg proteins, respectively. Furthermore, the presence of glycogen in liver section of USF2
/
mice was also
noted by histochemical staining (data not shown), indicating a normal
capacity of the liver for glycogen replenishment after refeeding.
Taken together, these results were indicative of an apparent normal glucose uptake and utilization by the liver of the knock-out mice.
In the USF2 /
mice presented in this paper, we have primarily
studied glucose metabolism and expression of glucose-responsive genes
in the liver. In the adult USF2
/
mice refed a high carbohydrate diet, we have assessed that glucose was available, normally
transported, and phosphorylated in hepatocytes. Indeed, in these mice,
glycemia was normal, the specific facilitative transporter GLUT2 that
provides glucose to the hepatocytes was present, and glucose
phosphorylation to glucose 6-P was normal as judged from the normal
accumulation of glucose 6-P in the liver of
/
mice.
Nevertheless, in the liver of these animals, we observed a significant
diminution of the L-PK and S14 gene expression 18 h after glucose
refeeding, whereas in ad libitum fed mice, no difference was
detected. This result, obtained in adult liver of USF2 /
mice, was
consistent with the 64% reduction in the amount of L-PK protein
noticed in the liver of
/
fetuses at the end of gestation, when the
L-PK promoter begins to be active (27). We conclude therefore that,
in vivo, USF2 proteins are required for a normal kinetics of
transcriptional induction of the two genes by glucose. This induction
is thought to depend on the nature of the glucose-response complex
assembled on the GlRE/ChoRE of the L-PK and S14. Although the exact
role of USF proteins in this complex remains disputed (28), evidence
argues for their important involvement, in cooperation with other still
unknown factors (19). Very recently, Kennedy et al. (29)
were able to demonstrate that microinjection of USF antibodies
specifically inhibited the glucose responsiveness of the L-PK promoter
in INS-1 rat insulinoma cells. Therefore, we carefully characterized
USF binding activity in the liver of animals refed a high carbohydrate
diet for 18 h. As previously reported in rat (14), USF1/USF2a
heterodimers are largely predominant in the liver nuclear extracts from
normal mice. In contrast, in nuclear extracts purified from the USF2
/
mice, the USF binding activity is exclusively accounted for by
USF1 homodimers, suggesting that no endogenous factor could replace
USF2 in
/
mice. Therefore, the question why USF1 homodimers are
less efficient than USF1/USF2 heterodimers in promoting glucose-induced
gene expression arises. Several mechanisms could be proposed.
USF1 homodimers could be in limiting concentration as compared with the heterodimers, consistent with a gene dosage effect. However, we were unable to detect any alteration of the L-PK and S14 gene expression in heterozygous mice.
Binding of USF1 homodimers could modify the binding of the other
transcriptional factors known to interact with the L-PK promoter (3).
This particular point was further documented by performing genomic
footprintings on liver cells of /
or +/+ refed
animals.2 In fact, the
results indicate that in vivo occupancy of the L-PK promoter
appears to be similar in USF2
/
mice and normal mice. In
particular, protection patterns on the GlRE are identical in
/
and
+/+ mice.
An alternative could be that USF dimers per se have different transactivating properties. Furthermore, since the NH2 activation domains of USF1 and USF2 are extremely different, they could specify differential interactions with other regulatory proteins (partners and/or coeffectors ... ). To enlighten the respective role of USF1 and USF2, and to assess whether they are involved in regulating different sets of genes, we have undertaken the knock-out of the USF1 gene.
When mice were fed for 5 days a high carbohydrate diet, as well as an
ad libitum regular diet, L-PK and S14 mRNAs content in
the liver of USF2 /
mice were not significantly different from USF2
+/+ mice. This result seems to indicate that the intrinsic transcriptional capabilities of glucose-responsive genes are normal in
USF2
/
mice but that kinetics of glucose-dependent
activation is slow. Alternatively, an altered transcriptional rate
could be compensated by progressive accumulation of stable mRNAs.
Indeed, we previously reported that L-PK mRNAs appear to be very
stable in carbohydrate refed animals (30).
In conclusion, we confirm in this paper that USF factors are involved
in the normal response of hepatic glucose-responsive genes to glucose
and that USF1 homodimers cannot totally replace the predominant
USF1/USF2 heterodimers normally present on the GlRE/ChoRE of the L-PK
and S14 genes. Further investigations will determine the respective
role of USF1 and USF2 in the transcriptional response to glucose. The
mechanisms of the other phenotypic abnormalities associated with
deficiency in these factors will be investigated in USF2 /
mice,
for instance, growth delay, relatively apathetic and increased
post-natal lethality which is not obviously related to the decreased
transcriptional activation by glucose in the liver.
We thank P. Chafey for expert technical assistance, B. Viollet for helpful advice, and M. A. Lefrançois-Martinez for providing the mouse S14 probe. We thank Martin Evans for the kind gift of CCB-ES cells.