(Received for publication, October 28, 1996)
From the Developmental Biology Research Centre, The Randall Institute, King's College London, 26-29 Drury Lane, London WC2B 5RL, United Kingdom
We have set out to test a model for tissue-specific gene expression that relies on the early replication of expressed genes to sequester limiting activating transcription factors. Using an erythroid cell line, we have tested the changes in the DNA binding activity of the lineage-restricted transcription factor GATA-1 through the cell cycle. We find that GATA-1 activity is low in G1, peaks in mid-S phase, and then decreases in G2/M. In contrast, the binding activities of two ubiquitous transcription factors, Oct1 and Sp1, remain high in G2/M. GATA-1 protein and mRNA vary in a similar manner through the cell cycle, suggesting that the expression of the gene or the stability of its message is regulated. Although a number of transcription factors involved in the control of the cell cycle or DNA replication have been shown to peak in S phase, this is the first example of a lineage-restricted transcription factor displaying S phase-specific DNA binding activity. One interpretation of these data leads to a model in which the peak in GATA-1 DNA binding amplifies the effect of early replication on the activation of erythroid-specific genes at the same time as preventing activation of non-erythroid genes containing GATA-responsive elements. These results may also relate to recent data implicating GATA-1 function in apoptosis and cell cycle progression.
The relationship between cell division and differentiation is still unclear. In the hematopoietic system, it has long been thought that differentiation requires prior or concomitant cell division (1), but direct evidence for a mechanistic relationship has been hard to obtain. One potential mechanism involves the effect of DNA replication on repressive chromatin and stable transcription complexes. The passage of a replication fork through a chromatin template is known to transiently disrupt both histone octamers and complexes of transacting factors required for programming gene expression (2, 3). Subsequently, as nucleosomes reform behind the replication fork, transcription factors have a window of opportunity to bind and either reprogram gene expression or reassemble committed genes (2, 4, 5).
Evidence in support of a role for DNA replication in gene activation
for a variety of viral and cellular genes has been provided by the
activating effect of an origin of replication in transient transfection
assays (6-9). Furthermore, in the normal cellular context, a strong
correlation exists between the replication timing of a gene and its
state of transcriptional activity. Specifically, replication early in S
phase is associated with expression for a large number of
tissue-specific genes in a variety of different cell types (10-13). A
well studied example is the globin locus, which is replicated early
in erythroid cells where it is transcribed but late in fibroblasts
where it is not (13, 14). The major determinant of erythroid expression
of this locus is the locus control region, a set of nuclease
hypersensitive sites located several kilobases upstream of the first
gene in the cluster. The locus control region is required both for
transcription and early replication of the
globin locus (15),
further maintaining the correlation between replication timing and
tissue-specific gene activation, leading to a model where early
replication removes repressive chromatin and allows activating
transcription factors to bind (4, 16). Transcription factor
availability is clearly an important component of such a model. If the
overall nuclear concentration of critical factors remained constant in
the cell cycle, titration of transcription factors by early replicating genes would limit their availability for binding to later replicating genes with the same control sequences, as has been suggested for the
early replicating somatic 5 S RNA genes of Xenopus (4, 17,
18). Alternatively, sequestration of transcription factors to prevent
activation of genes replicating in late S phase would be further
enhanced by a concomitant decrease in factor DNA binding activity.
These considerations have led to the suggestion that fluctuations in
the levels of critical transacting factors during S phase may amplify
the effect on chromatin remodeling conferred by early replication
(6).
We have set out to test this model, using an erythroid cell line that
contains both a transcriptionally poised globin gene that
replicates during the first half of S phase (14, 19) and GATA-1, a
lineage-restricted transcription factor crucial for erythroid
differentiation (20, 21) that has potential binding sites in the
regulatory sequences of all erythroid genes (22) and is expressed at
increasingly higher levels during erythroid maturation (23). We find
that the availability of GATA-1 but not ubiquitous factors does indeed
vary in the cell cycle in a manner likely to amplify the effects of
replication timing. GATA-1 therefore represents a candidate molecule
for linking erythroid transcription and early replication. These data
can also be linked with recent observations that GATA-1 may be directly
involved in control of the cell cycle and the choice among
proliferation, differentiation, and apoptosis (24-26).
MEL1 cells (27) were grown in RPMI 1640 supplemented with 2 mM glutamine and 10% fetal calf serum in 37 °C incubators containing 5% CO2. Asynchronous log phase cells were separated into different stages of the cell cycle by centrifugal elutriation, as follows. 108-109 cells were pelleted, washed in phosphate-buffered saline, resuspended at a concentration of approximately 107 cells/ml, and injected into the chamber of a Beckman JE-5.0 elutriation rotor spinning at 3000 rpm containing ice-cold phosphate-buffered saline pumped at 21 ml/min. The pump rate was incremented in 10-15 steps up to 65 ml/min, and elutriated fractions were collected onto ice. Fractions were then pelleted, resuspended in 5 ml of phosphate-buffered saline, and analyzed by Coulter counter to determine cell concentration and size distribution. Aliquots of 5 × 105 cells were removed, stained with propidium iodide (28), and analyzed by FACS to determine the quality of the elutriation. Adjacent fractions that consisted of essentially identical cell cycle profiles were pooled, and nuclear proteins were then extracted.
Cell ExtractsElutriated cells were pelleted and then taken
up at constant cell concentration in ice-cold 20 mM Hepes,
pH 7.0, 2 mM MgCl2 containing 0.05% Triton
X-100, and a mixture of protease (5 mM benzamidine, 5 mM phenylmethylsulfonyl fluoride, 2 µg/ml leupeptin, 2 µg/ml pepstatin, 2 µg/ml aprotinin, 5 µg/ml bestatin), kinase (0.5 mM EDTA), and phosphatase (1 M
-glycerophosphate, 0.2 M levamisole, 0.25 M
NaF, 20 mM sodium vanadate) inhibitors. After incubation on
ice for 15 min, lysed cells were pelleted at 6500 rpm in a microfuge
for 2 min at 4 °C, and the pelleted nuclei were resuspended in 2 volumes of 20 mM Hepes, pH 7.0, 2 mM
MgCl2 containing 350 mM NaCl plus inhibitors.
After 15 min on ice, nuclear debris was removed by centrifugation at
14000 rpm for 10 min. Glycerol was added to the supernatants to 25%,
and aliquots were flash frozen and stored at
70 °C.
Nuclear extracts (0.2-4 µl) were mixed with 7.5 µl of 4 × buffer (20 mM Hepes, pH 7.9, 2 mM MgCl2, 50 mM NaCl, 4% w/v Ficoll), 1 µg of poly(dIdC)·poly(dIdC) nonspecific competitor DNA, and specific competitor DNAs where appropriate. Specific 32P end-labeled oligonucleotides (GATA-1, GCAACTGATA AGGATTC (29); Oct1, AATTCACTGG TTCCCAATGA TTTGCATGCT CTCACTTCAC TG; Sp1, GCTAAGGCCC CGCCCCCACC AAGG (30); typically 20 fmols) were then added to a final volume of 30 µl, and the mixture was incubated on ice for 30 min. Binding reactions were then loaded onto a 4% non-denaturing polyacrylamide/0.5 × TBE gel that had been pre-electrophoresed at 200 V until the current across it was 11 mA. After 2 h electrophoresis at 4 °C, gels were dried down and autoradiographed. Laser densitometry was performed with a Molecular Dynamics Computing Densitometer.
The cell cycle position of the peak in GATA-1 DNA binding activity was
estimated by using the quantitative FACS data (see, for example, Fig.
1a). 2n and 4n, representing DNA
content at the beginning and end of the cell cycle, were defined as the
positions of the peaks in fractions 1 and 7, respectively. The
n value for the fraction with the highest GATA-1 DNA binding
activity was determined relative to these. Peak values from four
separate elutriations were used to derive a mean cell cycle position
for GATA-1 DNA binding activity.
Western Blotting
Cell equivalents of cell cycle-fractionated extracts were separated by polyacrylamide gel electrophoresis on 10% SDS-polyacrylamide gel electrophoresis gels and semi-dry electroblotted onto Hybond C+ nitrocellulose (Amersham). GATA-1 protein was immunodetected with anti-mouse GATA-1 antibody (a kind gift of J. D. Engel) after blocking with 5% milk powder in Tris-buffered saline containing 0.05% Tween 20, followed by binding with anti-rabbit IgG-horseradish peroxidase and treatment with ECL (Amersham).
Preparation of Total RNA and Northern BlottingTotal RNA was extracted from elutriated cells by the method of Chomczynski and Sacchi (31), separated on formaldehyde-agarose gels, alkaline blotted onto Hybond N+ membranes (Amersham), and probed with a 32P random hexamer-labeled GATA-1 XhoI fragment.
To perform an analysis of GATA-1 DNA binding activity through the cell cycle, asynchronous MEL cells (27) were fractionated by centrifugal elutriation. This technique separates bulk suspension cells by size and, because cell volume increases through the cell cycle, results in fractionation of cells according to their position in the cell cycle. The cell cycle composition of the fractionation was assessed by FACS analysis of propidium iodide-stained cells taken from each fraction (a typical example is shown in Fig. 1a). The percentages of cells in each phase of the cell cycle for a typical elutriation are summarized in Fig. 1b (cells are defined as G1, S, or G2/M depending on their DNA content (2n, 2-4n, and 4n, respectively)).
We assessed GATA-1 DNA binding activity by electrophoretic mobility
shift assay (EMSA). Nuclear proteins, salt extracted from equal numbers
of cells through the cell cycle, were incubated with an oligonucleotide
containing a strong GATA-1 binding site from the mouse 1 globin
promoter (29) and electrophoresed (Fig. 2a,
upper panel, lanes 1-7). Nuclear protein from
asynchronous MEL cells was also analyzed (lane 8), and the
identity of the GATA-1 band shift was confirmed by comparing the
effects of specific (self-competition, lane 9) and
nonspecific (an oligonucleotide containing a mutated GATA-1 site,
lane 10) oligonucleotide competitors. Clearly, there is
variation in GATA-1 DNA binding activity per cell through the cell
cycle, with low levels in G1 (lanes 1 and 2), an increase to a peak of activity in S phase (lane
5), followed by a decrease in late S/G2 (lanes
6 and 7). Comparison of the FACS profiles for these
fractions shows that GATA-1 DNA binding activity is highest in mid-S
phase cells (Fig. 1a, fraction 5) compared with
early S (Fig. 1a, fraction 4) and late S (Fig.
1a, fraction 6). Quantitation of the GATA-1 band
intensities indicates that there is a 4-fold increase in GATA-1 DNA
binding activity between early G1 and mid-S phase and an
11-fold decrease between mid-S and G2/M (Fig.
2b).
Analysis of GATA-1 DNA binding activity from several different elutriations revealed a consistent peak in mid-S phase. The position of this peak was quantitated by referring to the FACS analysis of each elutriation (see, for example, Fig. 1a). 2n and 4n, representing DNA content at the beginning and end of the cell cycle, were defined as the positions of the G1 and G2/M peaks in fractions 1 and 7, respectively. The n value for the GATA-1 binding peak in each case was calculated relative to these. Data from four separate elutriations yielded a mean value of 2.98n with a standard deviation of 0.04. Thus, GATA-1 binding activity reproducibly peaks in mid-S phase. The drop in GATA-1 DNA binding from mid-S to G2/M was also reproducible, averaging 7-fold in the four elutriations.
To determine whether the variation in GATA-1 DNA binding activity was particular to GATA-1 in this cell line or whether it was a general feature shared with other transcription factors, we determined the DNA binding activity of Oct1 and Sp1, two transcription factors whose cell cycle-regulated DNA binding activity has been characterized previously. Oct1 DNA binding activity has been studied in HeLa cells, and it does not appear to vary relative to bulk protein through the cell cycle except during mitosis, when it loses DNA binding activity (32, 33). We therefore determined Oct1 DNA binding activity in equal numbers of MEL cells by EMSA, using an oligonucleotide containing the octamer motif from the mouse IgH chain gene enhancer (Fig. 2a, middle panel). In contrast to the result seen with GATA-1, Oct1 DNA binding activity does not appear to peak in S phase but instead is highest in the last elutriated fraction, which contains a mixture of G2 and M cells (Fig. 2a, middle panel, lanes 1-7). As cell volume increases through the cell cycle, we conclude that the nuclear concentration of Oct1 competent for DNA binding remains broadly the same, consistent with the published data (32, 33). A second ubiquitous transcription factor, Sp1, has also been shown not to vary in DNA binding activity through the cell cycle in various cell types (34). We therefore sought to determine its activity in MEL cells. An oligonucleotide containing a binding site for Sp1 gave a similar result to Oct1, with the amount of Sp1 DNA binding activity increasing progressively through the cell cycle (Fig. 2a, lower panel). Thus, for two ubiquitously expressed transcription factors, it appears that there is an increase in DNA binding activity concomitant with increasing nuclear volume through the cell cycle, likely maintaining a broadly constant concentration of nuclear protein. In contrast, although GATA-1 DNA binding activity increases in a similar way through to S phase, it contrasts with the other two transcription factors by exhibiting a sharp decrease in activity in late S/G2. GATA-1 revealed a similar profile when we took into account the doubling in nuclear chromatin protein during the cell cycle. Fig. 2b is a plot of the data in Fig. 2a showing the change in DNA binding activities of GATA-1, Oct-1, and Sp1 for equal amounts of nuclear protein through the cell cycle. The DNA binding activities of all three factors increase between G1 and early S. However, GATA-1 DNA binding activity decreases in G2/M (to 9% of peak value), while Oct1 and Sp1 remain high in G2/M (93 and 147% of peak values, respectively).
We next asked whether the S phase peak in GATA-1 DNA binding activity
reflected a change in the amount of GATA-1 protein through the cell
cycle or whether GATA-1 DNA binding activity was being regulated by
some post-translational mechanism, such as phosphorylation. Nuclear
protein from equal numbers of cell cycle-fractionated cells was
separated by polyacrylamide gel electrophoresis, Western blotted, and
probed with an antibody specific for GATA-1 (Fig. 3b). Variation in the amount of GATA-1
protein through the cell cycle can be seen (lanes 1-7),
with a peak in mid-S phase (lane 5), in the same fraction
where DNA binding activity is at its highest (Fig. 3a,
lane 5). Furthermore, the magnitude of the changes in GATA-1
protein and DNA binding activity through the cell cycle is broadly
similar. Quantitation of band intensities indicates that between early
G1 and mid-S the amount of GATA-1 protein increases 5-fold
while DNA binding activity increases 4-fold. Similarly, between mid-S
and G2/M the amount of GATA-1 protein decreases 8-fold and
DNA binding activity 11-fold. Taken together, these data indicate that
the variation in GATA-1 DNA binding activity through the cell cycle is
primarily due to variations in the total amount of GATA-1 protein and
that post-translational mechanisms make at most a minor contribution to
the change in GATA-1 DNA binding activity.
The varying amounts of nuclear GATA-1 protein through the cell cycle might come about by changes in protein stability, subcellular localization, or gene expression. However, we have assayed GATA-1 DNA binding activity in both nuclear and cytoplasmic extracts and found similar S phase peaks with low levels in G1 and G2/M.2 In particular, we do not detect high levels of GATA-1 DNA binding activity in the cytoplasm in G2/M, which would be expected if the drop in nuclear GATA-1 DNA binding activity we observe in this phase of the cell cycle was due to shuttling between compartments. To investigate the possibility that the variation in the amount of GATA-1 protein was due to changing levels of gene expression, total RNA was extracted from cell cycle-fractionated cells, separated on an agarose gel, Northern blotted, and then probed for GATA-1 message (Fig. 3c). A peak in the level of steady state GATA-1 mRNA was seen in mid-S phase (lane 5), as found for both GATA-1 DNA binding activity and GATA-1 total protein (Fig. 3, a and b, lanes 5). The change in the amount of GATA-1 mRNA (6-fold) between early G1 and mid-S phase is similar to that of protein (5-fold) and DNA binding activity (4-fold). The drop in the level of message between mid-S and G2/M decreases only a little less (4-fold) than protein and mean DNA binding activity (8- and 7-fold, respectively). Thus, the primary control over GATA-1 DNA binding activity appears to be at the RNA level, but there may be some additional control at the protein level.
Variations in DNA binding activity through the cell cycle have been found for ubiquitous transcription factors involved in some of the central events of the cell cycle, including Jun (35), E2F (36), factors regulating the expression of genes involved in DNA synthesis (34, 35), and S phase-specific histone gene expression (19, 37). However, in contrast to these examples, we have now shown that GATA-1, a lineage-restricted transcription factor that targets tissue-specific genes, possesses S phase-specific DNA binding activity in cycling cells. Activation of GATA-1 expression has also been seen in quiescent hematopoietic progenitors upon entrance into the cell cycle (38).
There are several ways in which our observations can be interpreted.
One possibility is that GATA-1-directed control of the erythroid
program operates through the regulation of erythroid-specific replication origins containing GATA sites, causing early replication and hence gene activation. Since several transcription factors are
required as origin-specific replication factors by certain viruses,
including Oct1 and CCAAT/NF1 family members (39), it is possible
that specific replication origins located near to GATA-1-responsive
genes may be activated by GATA-1 early in S phase, inducing early
replication and subsequent transcriptional activation by GATA-1. An
origin of replication has been mapped between the
and
globin
genes of the human
globin locus (40), and it is the origin used for
early replication of this region in erythroid cells. Furthermore, there
is a requirement for the presence of the locus control region, which
contains several GATA-1 sites and directs
globin gene expression,
for early replication of this region of the chromosome (15). However,
there is little evidence supporting a role for GATA-1 in eukaryotic
origin activity or replication, although it does appear to cause growth
inhibition and bind to the origin of replication in Escherichia
coli (41). Furthermore, from our data the peak in GATA-1 DNA
binding activity (Fig. 2a) occurs too late to fire the
globin locus origin early in S phase (19).
A more attractive model based on the correlation between replication
timing of genes and their expression (2, 4, 5, 10) positions the effect
of GATA-1 DNA binding activity downstream of replication. We envisage
that the observed pattern of GATA-1 binding through the cell cycle may
be necessary to amplify the effect of early replication on
transcriptional activation. As early replicating erythroid genes become
stripped of chromatin and active gene complexes, GATA-1 binding sites
in regulatory elements, including the locus control region, become
available (Fig. 4, left side). By having
sufficiently high levels of GATA-1 to compete successfully with the
reassembly of repressive chromatin, the expression of GATA-1-activated
erythroid genes is maintained. We note that the replication of the
locus control region, the major determinant of erythroid expression in
MEL cells (14, 19), correlates well with the peak in GATA-1 DNA binding
activity.
High levels of GATA-1 throughout S phase could lead to the activation of later replicating genes that are not part of the erythroid program but are normally responsive to other members of the GATA family. Genes containing GATA sites are expressed in a wide range of non-erythroid tissues (22, 42-44). Given the high levels of GATA-1 found in MEL cells, it seems unlikely that passive titration of GATA-1 by binding sites in early replicating activated genes is sufficient to prevent the activation of later replicating genes, as has been suggested for the differential effect of transcription factor IIIA on the expression of somatic and oocyte 5 S RNA genes (4, 17). Instead, we suggest that the decrease in GATA-1 activity that occurs during late S phase is required to prevent activation of non-erythroid genes subject to activation by other GATA family members (Fig. 4, right side). By ensuring that GATA factor availability and target site accessibility occur in the same temporal compartment of S phase, subsets of genes with closely related control sequences can be activated independently.
Finally, another interpretation of these data has been prompted by several recent papers suggesting a role for GATA-1 in control of the cell cycle. Ectopic expression of GATA-1 in both hematopoietic and non-hematopoietic cell lines can lead to alterations in the length of segments of the cell cycle (24). In particular, high levels of ectopic GATA-1 cause S phase to lengthen in NIH3T3 cells, suggesting that the reduction in GATA-1 DNA binding activity we see here may be necessary for the correctly timed exit from S phase. GATA-1 also appears to function as an erythroid survival factor, since differentiation of GATA-1 negative ES cells along the erythroid lineage, or abrogation of GATA-1 activity in MEL cells, leads to premature apoptosis (25, 26). Thus, the high GATA-1 DNA binding activity we see in S phase may be necessary to prevent apoptosis. The emerging picture is that the choice among proliferation, differentiation, and apoptosis is affected by the level of GATA-1. Our data demonstrating a peak of GATA-1 DNA binding activity in mid-S, downstream of the G1/S cell cycle checkpoints, support the idea that GATA-1 has a cell cycle role in erythroid cells.
We thank Geoff Partington for assistance with some of the early EMSA experiments, Doug Engel for anti-GATA-1 antibody, Rob Nicolas for oligonucleotides, John Pizzey for assistance with analysis, and the ICRF FACS lab for help with cell cycle analysis. We are also grateful to Tariq Enver and Louis Mahadevan for critical reading of the manuscript.