Upregulation of Cell-Surface CD71 Marks the Onset of EpoR Dependence in Erythroid Progenitors
Mouse fetal liver between E11 and E15 is primarily an erythropoietic tissue. Cell surface markers CD71 and Ter119 may be used to identify differentiation-stage specific subsets, directly in primary tissue 
. Here we divided freshly harvested fetal liver cells into six CD71/Ter119 subsets that we termed S0 to S5 and that form a developmental sequence (). Cells isolated from subsets S1 to S5 show morphological features characteristic of erythroid maturation, including decreasing cell and nuclear size, nuclear condensation, and hemoglobin expression (). The precise proportion of fetal liver cells within each of the CD71/Ter119 subsets is a function of embryonic age, with the majority of cells being in the early, S0 and S1 subsets in E12. The more mature, S3 to S5 subsets are gradually populated with cells during subsequent embryonic days (E13 to E15) 
Upregulation of CD71 coincides with the onset of EpoR dependence and with S-phase of the last generation of CFU-e.
fetal liver is small and lacks morphologically identifiable hemoglobinized erythroblasts of the enucleated (definitive) lineage 
. Here we found that EpoR−/−
fetal liver does not contain subsets S1 to S5 (). This suggested that in the definitive erythropoietic lineage that gives rise to adult-type enucleated red cells, EpoR becomes essential on or prior to the transition from S0 to S1; subsets S1 to S5 are composed almost entirely of Epo-dependent erythroblasts. Of note, the small number (≈5%) of Ter119+
cells in the EpoR−/−
fetal liver are all nucleated erythrocytes of the transient yolk-sac (primitive) lineage (Figure S1A
The Majority of S0 Cells Are Erythroid Progenitors at the CFU-e Stage
Erythroid progenitors have traditionally been identified by their in vitro colony-forming potential. “Colony forming unit-erythroid” (CFU-e) are defined as cells that give rise to colonies containing 8 to 32 hemoglobinized cells after 2–3 days of in vitro culture in Epo 
. We investigated the colony-forming potential of cells sorted from each of the S0 to S3 subsets (). CFU-e potential was exclusive to S0 and S1 and was lost with the transition to S2. Cells in S2 and S3 gave rise to small, 2 to 4 cell clusters ().
The frequency of CFU-e obtained from sorted S0 cells was 65%–70% of the frequency from sorted S1 (). S1 consists entirely of Epo-dependent cells of similar maturation, with CFU-e potential (). Assuming similar plating efficiency for sorted S0 and S1 (of ≈30%, ), this suggested that CFU-e make up 65%–70% of the S0 subset. This is in agreement with our finding that fetal liver cells expressing non-erythroid lineage markers, which were limited to S0, formed up to 30% of this subset (, Figure S1B
). Non-erythroid colony-forming progenitors were also restricted to S0, where they formed less than 5% of all colony-forming cells (). Our conclusion that 65%–70% of S0 cells are CFU-e was further supported by single cell RT-PCR, which showed that 68% of S0 cells expressed EpoR mRNA (Figure S1C
In all the experiments that follow, “S0” refers to S0 cells from which cells expressing non-erythroid markers were excluded by flow-cytometric gating or sorting.
S1 Cells Are Synchronized in S-Phase of a Single Cell Cycle
To examine the cell cycle status of erythroid subsets S0 to S5 in vivo, we injected pregnant female mice with the nucleotide analogue bromodeoxyuridine (BrdU) and harvested fetal livers 30 min post-injection. We sorted cells from each of S0 to S5 and stained them with antibodies directed at BrdU (). Cells that incorporated BrdU were in S-phase of the cell cycle at the time of harvesting. Subsets S4 to S5 showed a rapid decline in the number of S-phase cells, consistent with cell cycle exit of terminally differentiating cells. Unexpectedly, we noted that ≈90% of S1 cells were in S phase, as compared with ≈50% of cells in S0 (). In addition, the intensity of the BrdU fluorescence within S1 cells was approximately 50% higher than in S0, suggesting a higher rate of DNA synthesis (Figure S1D
). Similar experiments with EpoR−/−
fetal liver showed that EpoR appears to have no effect on progenitor cell cycle status (Figure S1E
Consistent with the higher number of S-phase cells in S1, we found a corresponding increase in the E cyclins in S1 compared with S0 (). Strikingly, we noted >30-fold decrease in the CDKI p57KIP2
mRNA, but no significant change in the mRNA of other members of the CIP/KIP CDKI family; there was induction in p27KIP1
later in differentiation, in subsets S2 and S3 ( and Figure S1F
. The p57KIP2
protein also decreased at the S0 to S1 transition ( lower panel).
The finding that nearly all S1 cells were in S-phase could be due to an unusual cell division cycle with short or no gap phases. Alternatively, S1 cells may be synchronized in S-phase of the cycle. The latter explanation would require that cells spend only a brief period of a few hours in S1, lasting through part or all of a single S phase. The preceding G1 phase of this same cell cycle would have occurred prior to the transition from S0 to S1. The G2 and M phases of this same cycle would occur as cells upregulate Ter119 and transition into S2.
To investigate these possibilities, we isolated S0 cells by flow-cytometry, labeled them with the cell-tracking dye carboxyfluorescein diacetate succinimidyl ester (CFSE), and followed their Epo-dependent differentiation into S1 in vitro (). By 10 h, 53% of S0 cells transitioned into S1 in the absence of cell division, as indicated by a single CFSE peak for S1 (solid red histogram, t
10 h) that was identical in intensity to that of the CFSE peak for S0 (blue histogram, t
10 h; median CFSE fluorescence for both S1 and S0 peaks
4,400). This suggested that the transition from S0 to S1 occurred in the absence of cell division, within a single cell cycle. Four hours later, at t
14 h, essentially all S1 cells had divided once, as indicated by the halving of the CFSE signal (red histogram at t
14 h, CFSE fluorescence
2,100). The simultaneous division of S1 cells suggested they were synchronized in their cell cycle phase. By contrast, only a portion of S0 cells, which were presumably asynchronous in their cell cycle phase, had divided at this time, resulting in a biphasic CFSE peak (blue histogram, t
Taken together, these results suggest that the most mature CFU-e progenitor (“CFU-e.2”, ), capable of giving rise to an eight-cell colony, traverses S0, S1, and enters S2 within a single cell cycle. This progenitor arises in S0, becomes Epo dependent, and upregulates CD71, transitioning into S1 during S-phase of its cell cycle. Upregulation of Ter119 occurs at approximately the same time that it completes its cycle and divides, giving rise to progeny that lack CFU-e activity in S2 (). These conclusions are consistent with essentially all S1 cells being in S-phase (), and with our finding that nearly all S1 cells are sensitive to hydroxyurea, a drug that specifically targets S-phase cells (Figure S2A
). These conclusions are consistent with a number of other observations: the loss of CFU-e activity with Ter119 expression (, 
), the short time span (<15 h) that freshly sorted S0 cells require to transition through S1 and into S2 (compare with an estimated cell cycle length of 16 h for a CFU-e cell that will undergo three cell divisions in 48 h, giving rise to an eight cell colony), and with early work suggesting that Epo dependence first occurs in early S-phase of a specific CFU-e cell generation 
. These conclusions are also consistent with the finding that EpoR−/−
embryos have normal numbers of CFU-e 
: though EpoR−/−
embryos lack S1 cells, all the CFU-e in S1 first arise as Epo-independent cells in S0, where they are presumably retained in the EpoR−/−
S-Phase Progression Is Required for the Transition from S0 to S1
There are two ways to explain how upregulation of CD71, a differentiation event, might coincide with S-phase, a cell cycle event. These events may have each been initiated in parallel by a common upstream regulator, such as the EpoR, since both occur at the time that cells become EpoR dependent. Alternatively, there may be a direct mechanistic link between the differentiation and cell cycle programs. To distinguish these possibilities, we examined whether a block to S-phase progression would interfere with CD71 upregulation (). We incubated sorted S0 cells in vitro for 10 h in the presence of Epo, and either in the presence or absence of aphidicolin, an inhibitor of DNA polymerase that arrests S-phase progression 
. At t
10 h, cells were washed free of aphidicolin and incubated in Epo alone for an additional 10 h (). In the initial 10 h of incubation, there was an Epo-dependent transition of cells from S0 to S1 (, rows 1 and 5). However, the presence of aphidicolin blocked this transition (, rows 2 & 3, t
10 h). Both S-phase and the transition into S1 resumed once the cells were washed free of aphidicolin (, rows 2 & 3, t
20 h). These observations suggested that the transition from S0 to S1 occurred during S-phase and required both Epo and S-phase progression.
The S0 to S1 transition requires S-phase progression.
We also examined the effect of mimosine, a plant amino acid that blocks cell cycle progression in late G1 
. We incubated sorted S0 cells in Epo and in the presence or absence of mimosine. By 4 h of incubation, the majority of cells were arrested in G1. However, a small fraction of cells (12%) could be seen in S-phase at t
4 h (, row 4, BrdU/7AAD at t
4 h). Presumably, at the time mimosine was added, these cells were advanced in their cell cycle beyond the point at which mimosine exerts its block. BrdU/7AAD analysis showed that these cells were in the early half of S-phase and expressed the highest CD71 levels within the S0 subset (, cells marked in red). By t
10 h, no S-phase cells were seen in S0, presumably because they have now transitioned into S1, where a similar number of cells (15%) had newly appeared (, row 4, BrdU/7AAD for S0 at t
10 h, and CD71/Ter119 for S1 at t
10 h). These observations were consistent with the onset of CD71 upregulation occurring in early S-phase in S0, culminating in the transition to S1 later within that same S-phase.
CD71, the transferrin receptor, is required during erythroid differentiation in order to facilitate cellular uptake of iron for hemoglobin synthesis. CD71 is also expressed, albeit at lower levels, on all cycling cells. We therefore examined whether, in the context of S1 cells, CD71 might be required specifically for S-phase progression. We used RNAi to prevent CD71 upregulation in S0 cells during their incubation in Epo (Figure S2B
,C). The failure of these cells to upregulate CD71 did not interfere with the number of cells in S-phase (Figure S2B
). Therefore, the link between S-phase progression and CD71 upregulation in S1 cells is not due to a cell cycle function for this gene.
The S0 to S1 Transition Is Marked by Downregulation of PU.1 and GATA-2 and Precedes Induction of Erythroid-Specific Genes
To investigate the link between S-phase and the erythroid differentiation program, we examined expression of erythroid transcriptional regulators and erythroid-specific genes in freshly sorted fetal liver subsets and in fetal brain (). We found that the GATA-1 mRNA was present in S0 cells, at 200-fold higher levels than in fetal brain () and 40-fold higher level than in Mac-1+
cells (Figure S3A
). It increased a further ≈2-fold with the transition from S0 into S1 and continued to increase in S2 and S3. Of note, total RNA per cell decreased 4-fold over the course of differentiation from S2 to S4 (Figure S3B
), suggesting an overall modest increase in GATA-1 mRNA per cell over this period. Other erythroid transcriptional activators and GATA-1 associated factors, including EKLF, NF-E2 
, SCL/Tal-1, and Lmo2, showed a similar expression pattern to that of GATA-1 (). Therefore, expression of GATA-1 and of other activators of the erythroid transcriptional program precedes the transition from S0 to S1. By contrast, we found that PU.1, a repressor of GATA-1 function, and GATA-2, a target of GATA-1-mediated repression 
, were both downregulated ≈30-fold and ≈20-fold, respectively, at the S0 to S1 transition, becoming undetectable with further differentiation (). Prior to its downregulation, the level of PU.1 in S0 cells was comparable to that of myeloid Mac-1+
cells (Figure S3A
). PU.1 protein levels also declined with the transition from S0 to S1 (Figure S3C
Block of S-phase progression at the S0 to S1 transition arrests the erythroid differentiation program.
EpoR−/− fetal liver cells, though apparently arrested at the S0 stage (), have a similar expression pattern of transcriptional regulators to wild-type S1 (). Therefore, downregulation of PU.1 and GATA-2 at the S0 to S1 transition, as well as the preceding induction of GATA-1, are independent of EpoR signaling.
We examined expression of several erythroid-specific GATA-1 target genes: β-globin (Hbb-b1
); the first enzyme of heme synthesis, aminolevulinic acid synthase 2 (ALAS2); and the anion exchanger Band 3 (Slc4a1
), a major erythrocyte membrane protein 
. There was a modest increase in their expression at the S0 to S1 transition, followed by a 30–100-fold induction during subsequent differentiation in S2 and S3 (). Expression of the EpoR gene, itself a GATA-1 target, increased 10-fold above its S0 level with the transition to S1 (Figure S3D
). Taken together, induction of erythroid GATA-1 target genes and repression of GATA-2 suggest that GATA-1 function is activated at the S0 to S1 transition. The modest increase in GATA-1 mRNA at this time suggests that its activation may be principally a result of PU.1 downregulation.
S-Phase Arrest at the S0 to S1 Transition Blocks Induction of Erythroid-Specific Genes
We had found that S-phase progression at the transition from S0 to S1 was required for CD71 upregulation (). We therefore examined whether S-phase progression at this time was also required for induction of erythroid-specific genes. We cultured sorted S0 cells in Epo for 10 h, a period sufficient for 25%–50% of cells to transition into S1 (, ), and examined the effect of adding aphidicolin to the culture. Cells were then washed free of aphidicolin, continuing incubation in Epo alone. Cells incubated in Epo alone for the entire period showed ≈50- to 100-fold induction in the mRNAs for β-globin, Band 3, and ALAS2 (, red curves). By contrast, cells that were subject to aphidicolin treatment during the initial 10 h showed reduced mRNA induction by the end of the culture period (, blue curves). The reduced mRNA levels corresponded closely to the levels predicted had there been a 10 h delay in the time course of induction for each of the genes (, black curves). Therefore, induction of erythroid-specific genes was likely blocked during the incubation period in aphidicolin.
We also examined whether S-phase arrest interferes with erythroid gene induction if applied at the S1 stage of differentiation. We sorted S1 cells and incubated them in Epo, either in the presence or absence of aphidicolin. Unlike S0 cells, aphidicolin-mediated S-phase arrest of S1 did not interfere substantially with their induction of erythroid specific genes, as shown by the unperturbed induction of β-globin, Alas2, and Band 3 (, Figure S3E
) or with the upregulation of Ter119 (Figure S3F
). Therefore, S-phase progression is required for activation of erythroid-specific genes, specifically at the S0 to S1 transition, but not a few hours later when the cells have traversed into S1. The lack of effect of aphidicolin on mRNA induction in S1 suggests its effects in S0 are not due to non-specific suppression of transcription.
S-Phase Arrest at the S0 to S1 Transition Blocks Downregulation of PU.1 and GATA-2 and Arrests Erythroid Morphological Maturation
Transcripts for PU.1 and GATA-2 are markedly downregulated at the transition from S0 to S1 (). We examined whether S-phase arrest interferes with their downregulation. Sorted S0 cells were incubated in Epo for 4 h, at which time, just prior to their transition into S1 (), aphidicolin was added to the cultures for a period of 10 h. Cells were then washed free of aphidicolin and incubated in Epo for a further 10 h. Aphidicolin halted the downregulation of both PU.1 and GATA-2, which resumed once the cells were washed free of the drug (). Similar results were obtained in cells treated with mimosine (Figure S3G
). Therefore, S-phase progression is required for downregulation of PU.1 and GATA-2 at the S0 to S1 transition. Of note, GATA-1, Nfe2, and Lmo2 mRNAs, which did not change significantly during the transition from S0 to S1 (), were not altered significantly by the aphidicolin treatment (, Figure S3H
We also examined the effects of aphidicolin or mimosine treatment on morphological maturation of S0 cells cultured in Epo. Following 10 h in Epo in the presence of aphidicolin or mimosine, cells appeared larger than cells incubated in Epo alone. This suggested that, while S-phase progression and the erythroid differentiation program had both arrested, cell growth was not perturbed (). Cells were then washed free of aphidicolin or mimosine and cultured in Epo alone. By 20 h, erythroid maturation had resumed in cells that were initially incubated in cell cycle blocking drugs, as judged by decreasing cell size, nuclear condensation, and decreased nuclear to cytoplasmic ratio, but was nevertheless delayed when compared with control cells. These results are consistent with the effect of S-phase arrest on gene expression () and suggest that S-phase progression at the S0 to S1 transition is a key requirement for activation of the erythroid differentiation program.
Preventing p57KIP2 Downregulation Blocks S-Phase Progression at the S0 to S1 Transition and Arrests Erythroid Differentiation
Expression of p57KIP2
mRNA decreases over 30-fold at the S0 to S1 transition, and this is associated with downregulation of the p57KIP2
protein (). To examine the effect of preventing p57KIP2
downregulation, we generated a point mutant of p57KIP2
, p57T329A, analogous to a proteolysis-resistant human p57KIP2
. Sorted S0 cells were infected with bicistronic retroviral vectors expressing either wild-type p57KIP2
or p57T329A, linked through an internal ribosomal entry site (IRES) to a human CD4 (hCD4) reporter; control cells were infected with retroviral vector expressing the IRES-hCD4 construct only (MICD4). To allow expression of the transduced p57KIP2
, infected cells were cultured for 15 h in stem-cell factor (SCF) and interleukin 3 (IL-3), cytokines that sustain viability of progenitors but, unlike Epo, do not support differentiation from S0 to S1. Infected S0 cells were then transferred to Epo for 14 h (). Expression of either wild-type (unpublished data) or mutant p57KIP2
, but not expression of MICD4, resulted in a block to S-phase progression and inhibited the transition from S0 to S1 (). Further, PU.1 mRNA was >3-fold higher in cells expressing p57KIP2
compared with control cells expressing vector only (), suggesting that, as in the case of aphidicolin-mediated S-phase arrest, p57KIP2
-mediated S-phase arrest prevents downregulation of PU.1 at the transition from S0 to S1. Erythroid morphological maturation, but not cell growth, of p57T329A-transduced cells was also arrested (Figure S3I
Taken together, upregulation of CD71, which defines the transition from S0 to S1, identifies a key differentiation transition within the last generation of CFU-e (“CFU-e.2”, ). It marks the onset of EpoR dependence and occurs exclusively during S-phase of the cell cycle. Induction of GATA-1 and other activators of the erythroid transcriptional program precedes this transition, whereas induction of erythroid-specific genes such as β-globin and Ter119 follows it. The S0 to S1 transition coincides with rapid downregulation of p57KIP2
, PU.1, and GATA-2. Both Epo and S-phase progression are required for upregulation of CD71. S-phase progression at the S0 to S1 transition requires the downregulation of p57KIP2
and is in turn required for the downregulation of PU.1 and GATA-2 and the subsequent activation of erythroid-specific genes. By contrast, S-phase arrest in S1 cells does not affect erythroid gene activation (, S3E–F
Persistently Elevated PU.1 Arrests S-Phase Progression and Blocks Erythroid Differentiation
Both PU.1 and GATA-2 were rapidly and dramatically downregulated at the transition from S0 to S1 (, Figure S3A
). We examined the effect of preventing this downregulation by expressing either PU.1 () or GATA-2 (, S4C,D
) in S0 cells using retroviral constructs and a similar strategy to that described above for p57KIP2
. Following infection, S0 cells were cultured for 15 h in IL-3 and SCF and then transferred to Epo for 24 h, when CD71/Ter119 and cell cycle profiles were examined (). We divided the PU.1 expression profile at t
24 h into 7 sequential hCD4 gates labeled (i) to (vii) (), each containing cells with increasing levels of the hCD4 reporter and, therefore, increasing levels of PU.1. By measuring PU.1 protein directly in fixed and permeabilized cells using a PU.1-specific antibody and flow-cytometry, we found that hCD4 protein expression was a reliable reporter of exogenous PU.1 protein expression in our system (, S4A–B
); expression of transduced PU.1 was also measured by qPCR (Figure S4D
). Sequential hCD4 gates were also obtained for control cells expressing the empty MICD4 vector. PU.1 expression blocked transition from S0 to S1, with the number of cells transitioning into S1 declining as PU.1 expression increased (, upper panels). PU.1 expression also resulted in a decrease in the number of S-phase cells, with cells arresting principally at the transition from G1 to S-phase, though there was also an increase in the number of cells within G2 or M (, lower panels). The decrease in the number of cells in S1 was paralleled by decreased S-phase cell number, suggesting a direct correlation between the PU.1-mediated block of the transition from S0 to S1, and its inhibitory effect on S-phase (). Therefore, PU.1 inhibits both S-phase and erythroid differentiation at the S0 to S1 transition.
PU.1, but not GATA-2, inhibits the transition from S0 to S1.
Since the downregulation of both PU.1 and p57KIP2
are required for S-phase progression and for the transition from S0 to S1 (, ), we examined whether PU.1 may be a regulator of p57KIP2
. However, we found that exogenous expression of PU.1 did not prevent downregulation of p57KIP2
). Therefore, PU.1's inhibitory effect on S-phase is not mediated via p57KIP2
In contrast to PU.1, expression of GATA-2 in S0 cells did not prevent transition into S1, though it somewhat reduced the subsequent transition from S1 to S2 (). GATA-1 overexpression in S0 cells had the opposite effect, of promoting the transition from S1 to S2. There was no significant effect of either GATA-1 or GATA-2 on the cell cycle profile ().
The S0 to S1 Transition Coincides with a Switch in the Timing of Replication of the β-Globin Locus
A long-standing hypothesis suggests that DNA replication may provide an opportunity for the restructuring of chromatin at tissue-specific gene loci 
. Given the requirement for DNA replication for the transition from S0 to S1, we asked whether chromatin change may be taking place at this time. The β-globin gene locus () is a well-studied model of tissue-specific gene expression. The features that characterize the open chromatin conformation at the actively transcribed locus in erythroid cells have been established, but the time during development when the active chromatin conformation is acquired is not known. We therefore set out to examine whether the S0 to S1 transition might coincide with an alteration in the structure or function of chromatin at this locus.
The S0 to S1 transition coincides with an S-phase dependent switch in the state of chromatin at the β-globin locus.
The timing of replication of the β-globin locus is correlated with its chromatin state. In higher eukaryotes the timing of replication of genes correlates with their transcriptional activity 
. Housekeeping genes replicate early in S-phase, whereas silent chromatin and heterochromatin replicate late. The β-globin locus replicates in mid to late S-phase in non-erythroid cells, and early in S-phase in erythroid cells 
. We examined the timing of replication of the β-globin locus in S0 and S1 cells sorted from fresh fetal liver. Individual alleles were identified using fluorescence in situ hybridization (FISH) with a probe directed at the β-major gene. Cells in S-phase were identified by positive staining for BrdU incorporation. Nuclei from at least 100 S-phase cells from either S0 or S1 were examined in each of two experiments (). Using this approach, two single dots (“SS”) suggest that neither of the β-globin alleles had yet replicated. Nuclei in which both alleles have replicated contain a pattern of two double dots (“DD”). Replication of only one allele results in one single and one double dot (SD) 
. We found that the number of cells with a DD pattern increased from only 15% in S0 to over 50% in S1 (), suggesting a switch in the timing of replication from late to early S-phase. In addition, an average of 36% of S0 cells, but only 21% of S1, had an SD pattern, consistent with a switch from late, asynchronous replication in S0 to early, synchronous replication in S1 
The S0 to S1 Transition Coincides with the Onset of DNase I Hypersensitivity at the β-Globin Locus Control Region (LCR)
A key indicator of open chromatin at the β-globin LCR is the presence of hypersensitivity (HS) sites (). We prepared nuclei from freshly sorted S0 or S1 cells and tested their sensitivity to DNase I digestion. Following digestion, we measured remaining DNA using quantitative PCR, with amplicons within HS2, HS3, and HS4 
. Results were expressed as a ratio to the DNase I resistant, non-expressing neural gene, Nfm. We found that S0 cells were relatively resistant to DNase I, while S1 cells were hypersensitive at all tested HS sites (). Therefore, the S0 to S1 transition coincides with the onset of DNase I hypersensitivity at the β-globin LCR.
We also examined E12.5 EpoR−/− whole fetal livers, which do not contain S1 cells (). We found that EpoR−/− fetal livers were resistant to DNase I, whereas whole fetal livers from wild-type or heterozygous littermates showed the expected hypersensitive sites (). We therefore concluded that DNase I hypersensitivity develops at the S0 to S1 transition, synchronously with the onset of EpoR dependence.
S-Phase Progression Is Required for the Onset of DNase I Hypersensitivity at the β-Globin LCR
Since the transition from S0 to S1 coincides with, and requires, S-phase progression, we examined whether development of DNase I hypersensitivity at the β-globin LCR also requires S-phase progression. We incubated sorted S0 cells in Epo, in the presence or absence of aphidicolin, for 10 h. Over this period 25%–50% of S0 cells transition into S1, a process arrested by aphidicolin (, ). At the end of a 10-h incubation period, nuclei were prepared and digested with varying concentrations of DNase I. There was a clear increase in DNase I sensitivity in cells incubated in Epo alone, relative to cells incubated in Epo and aphidicolin (). Therefore, the development of DNase I hypersensitivity at the S0 to S1 transition is dependent on S-phase progression.
Changes in Post-Translational Histone Tail Modifications Associated with the Transition from S0 to S1
The switch in timing of replication and in DNase I hypersensitivity at the S0 to S1 boundary suggested the β-globin LCR was undergoing structural changes. To investigate these, we used chromatin immunoprecipitation (ChIP) to determine specific histone tail modifications at the β-globin LCR in freshly sorted S0, S1, and in fetal brain. We used ChIP-qPCR for amplicons at the β-globin LCR HS sites, or at a control, neural gene, Nfm. Changes in histone modifications were expressed as a ratio, between S0 and either S1 or fetal brain (). summarizes data pooled from seven experiments with various immunoprecipitating antibodies as indicated. A comparison of S1 with S0 shows a 7-fold decrease in trimethylation of histone 3 lysine 27 (H3K27me3, p
0.019, paired t
test), a mark associated with silent chromatin, and a 2.5-fold increase in histone 3 lysine 4 dimethylation, a mark associated with active chromatin (H3K4me2, p
0.032), at the HS2 site of the β-globin LCR. A similar trend for these two modifications was also found at other HS sites (p
0.0006 and p
0.011 for H3K4me2 and H3K27me3, respectively, pooling all HS sites). An increase in acetyl marks in histones H3 and H4 associated with active chromatin was also seen consistently across the HS sites tested, though it did not reach statistical significance. Of note, no significant changes in histone marks were found between S0 and S1 at the Nfm gene. Further, there was no significant change in total histone occupancy of the HS sites between S0 and S1, as determined by ChIP with antibodies directed against total H3 and H4 ().
We noted that H3K27me3, associated with silent chromatin, and H3K4me2, associated with active chromatin, were both enriched in S0 compared with fetal brain (, lower panel). These results were suggestive of bivalent chromatin at the β-globin LCR in S0, and loss of the repressive H3K27me3 mark with transition into S1 (, upper panel, 5F).
The Transition from S0 to S1 Coincides with S-Phase-Dependent DNA Demethylation at the β-Globin LCR
We examined DNA methylation of six CpG dinucleotides, three each at the HS1 and HS2 sites of the β-globin LCR (, ). Genomic DNA was prepared from sorted hematopoietic cell subsets from fresh fetal liver, including S0, S1, megakaryocytic CD41+, myeloid Mac-1+, and Lin−Sca1+Kit+ (LSK) cells, enriched for hematopoietic stem-cells. We also examined EpoR−/− fetal livers depleted of cells expressing lineage markers, and fetal brain. DNA methylation at each of the six CpGs was obtained following bisufite conversion of genomic DNA, PCR amplification at HS1 and HS2, and pyrosequencing. In fetal brain methylation levels were high, at ≈60%–80%, for all six CpG dinucleotides. Methylation levels were lower in all hematopoietic cell subsets (). Methylation levels were largely similar in all hematopoietic, Epo-independent cell subsets examined: LSK, Mac-1+, CD41+, S0, and EpoR−/− cells. The onset of Epo dependence in S1 was associated with a marked reduction in DNA methylation in all six CpG dinucleotides, with the level of methylation dropping to virtually undetectable levels in S1 for four of the six CpGs.
The transition from S0 to S1 is marked by the onset of S-phase dependent, DNA demethylation at HS1 and HS2.
We found that DNA demethylation also took place in freshly sorted S0 cells allowed to differentiate in vitro (). Demethylation in vitro occurred earlier at the HS1A, B, C, and HS2C than at HS2A, B (, red lines), in agreement with results in vivo (). Demethylation in vitro was arrested at all CpGs if either aphidicolin or mimosine were added to the incubation medium, and resumed when these drugs were removed (). Therefore, DNA demethylation, initiated at the transition from S0 to S1, is dependent on S-phase progression. These results are suggestive of a passive demethylation process, due to loss of maintenance methylation at nascent DNA.
PU.1 Downregulation Is Required for the Onset of DNA Demethylation at the Transition from S0 to S1
We examined HS1 and HS2 DNA methylation levels in S0 cells transduced with PU.1 (as in ) and incubated in Epo for 24 h. DNA methylation was significantly higher at 3 of the 6 CpGs in S0 cells transduced with PU.1-ICD4, compared with control cells transduced with MICD4 (). Therefore, PU.1 expression, along with its inhibitory effect on erythroid differentiation, also impaired DNA demethylation, possibly due to its inhibitory effect on S-phase in these cells ().