Replication stress caused by low MCM expression limits fetal erythropoiesis and hematopoietic stem cell functionality

Replicative stress during embryonic development influences ageing and predisposition to disease in adults. A protective mechanism against replicative stress is provided by the licensing of thousands of origins in G1 that are not necessarily activated in the subsequent S-phase. These ‘dormant' origins provide a backup in the presence of stalled forks and may confer flexibility to the replication program in specific cell types during differentiation, a role that has remained unexplored. Here we show, using a mouse strain with hypomorphic expression of the origin licensing factor mini-chromosome maintenance (MCM)3 that limiting origin licensing in vivo affects the functionality of hematopoietic stem cells and the differentiation of rapidly-dividing erythrocyte precursors. Mcm3-deficient erythroblasts display aberrant DNA replication patterns and fail to complete maturation, causing lethal anemia. Our results indicate that hematopoietic progenitors are particularly sensitive to replication stress, and full origin licensing ensures their correct differentiation and functionality.

T he process of genomic duplication starts at replication origins, which are licensed in the G1 phase of the cell division cycle, several hours before their activation in S phase. The licensing process is led by the origin recognition complex (ORC), cell division cycle 6 (CDC6) and Cdc10dependent transcript 1 (CDT1) proteins, which cooperate to engage the mini-chromosome maintenance (MCM) complex with the DNA. MCM, composed by essential subunits MCM2-7, displays DNA helicase activity and becomes part of the replisome machinery (reviewed in references 1,2 ) . Defective control of DNA replication causes 'replicative stress' (RS), which is the underlying cause of several developmental diseases. Mutations in ORC, CDC6 and CDT1 genes are related to Meier-Gorlin syndrome, a type of dwarfism [3][4][5] , and mutations in MCM4 are linked to growth retardation, adrenal insufficiency and natural killer cell deficiency 6 . Impaired MCM function also increases cancer susceptibility [7][8][9][10][11][12][13] (reviewed in references 14,15 ).
MCM complexes are normally loaded onto DNA in excess relative to the number of origins that fire during the S phase (reviewed in reference 16 ). One function of the surplus of MCM is to license dormant origins that may be activated in response to stalled or collapsed forks, providing a rescue mechanism under RS [17][18][19] . Another possible function for the high number of licensed origins, which remains largely underexplored, is to provide flexibility to the replication process during early embryonic development 20,21 or in cell differentiation contexts that require the activation or shut-off of specific origins [22][23][24] (reviewed in references 25,26 ).
To investigate the protective effects of MCM against RS in vivo, a new Mcm3 conditional knockout (KO) mouse model was designed. The targeted allele, Mcm3-Lox, was hypomorphically expressed and caused embryonic lethality in homozygosity, which could be partially alleviated in a genetic background with enhanced resistance to RS. Our results show how a full complement of MCM proteins is specifically required to preserve the functionality of hematopoietic stem cells (HSCs) and the proper differentiation and maturation of erythrocytes in the developing embryo.

Results
A hypomorphic Mcm3 allele causes embryonic lethality. A modified mouse Mcm3 allele was designed with loxP sites flanking exons 14-17 and a luciferase reporter inserted at the 3 0 UTR under the control of an IRES element. The resultant allele (Mcm3-Lox) was intended as a conditional KO, as Mcm3 expression could be ablated with Cre recombinase (Fig. 1a and Supplementary Fig. 1A,B). In addition, expression of Mcm3-Lox could be monitored by the bioluminescence activity associated to luciferase expression. Mcm3 þ /Lox mice were born at Mendelian rates, and Mcm3-Lox expression in the skin and several internal organs was confirmed by luciferase activity after the administration of luciferin ( Fig. 1b and Supplementary Fig. 1C). Unexpectedly, virtually no Mcm3 Lox/Lox individuals were born after extensive breeding of Mcm3 þ /Lox mice (1 in 314; Fig. 1c). Mcm3 Lox/Lox embryos survived until E16.5-E18.5 but were noticeably smaller than wild-type (wt) or Mcm3 þ /Lox littermates and their pale appearance suggested an anemic phenotype (Fig. 1d). Mcm3 þ /Lox and Mcm3 Lox/Lox embryonic fibroblasts (MEFs) could be obtained and maintained in culture.
To determine the impact of the Mcm3-Lox modification on gene function, mRNA isolated from MCM3 Lox/Lox MEFs was sequenced to rule out the presence of any mutation that could affect MCM3 protein functionality ( Supplementary Fig. 2 (Fig. 1e), and consistent reductions in MCM3 protein were observed in whole cell extracts ( Fig. 1f; compare lanes 3-5-7). These results indicate that the Mcm3-Lox allele is hypomorphically expressed, presumably due to the modifications introduced in the 3 0 UTR. This level of reduction in MCM3 did not significantly affect the cellular concentration of other MCM subunits (Fig. 1f).
At least in primary embryonic fibroblasts, downregulation of Mcm3 expression was compatible with DNA replication and cell proliferation ( Supplementary Fig. 3A,B). Analyses of inter-origin distances (IOD) using stretched fibers revealed a similar frequency of origin activity between Mcm3 þ / þ and Mcm3 Lox/Lox MEFs in normal growth conditions (median IOD values of 118 and 116 Kb, respectively; Supplementary Fig. 3C). Upon challenging fork progression with aphidicolin, the IOD was shortened by 37% in Mcm3 þ / þ MEFs, as a consequence of the activation of dormant origins (median IOD 73.8 Kb; Supplementary Fig. 3C). In Mcm3 Lox/Lox MEFs, the reduction in IOD was limited to 24% (median IOD 87.5 Kb), reflecting a partial loss of back-up origins but also underscoring the large excess of origins licensed in G1, as even after the marked reduction in MCM3 protein concentration, a reservoir of additional origins were activated when needed. After several days in culture, Mcm3 Lox/Lox MEFs displayed gH2AX foci, a common marker of DNA damage associated to RS ( Supplementary Fig. 3D).
Mcm3-deficient mice are prone to haematological neoplasia. In contrast to Mcm3 Lox/Lox , heterozygous Mcm3 þ /Lox mice were viable and could be crossed with a CMV-Cre strain expressing Cre recombinase to generate viable Mcm3 þ / À mice. As expected, breeding between Mcm3 þ / À mice did not produce any Mcm3 À / À offspring (Fig. 1g). All previous attempts at making MCM-null models displayed pre-implantation lethality [8][9] . When cohorts of Mcm3 þ / À and Mcm3 þ /Lox mice were established for longevity studies, both groups presented a slight but significant reduction in lifespan (Fig. 1h) and an increased incidence of tumors, mainly lymphomas originated in the mesenteric lymph nodes that in multiple cases infiltrated to other organs. These tumors are not uncommon in aged mice but normally have a later onset and lower invasiveness (Supplementary Tables 1-3).
DNA damage and incomplete erythropoiesis in the fetal liver. We next focused on the tissues that might be responsible for the lethality observed in Mcm3 Lox/Lox embryos, which by mid-late gestation (E14. 5-18.5) were consistently smaller than their Mcm3 þ / þ or Mcm3 þ /Lox counterparts. In control E16.5 embryos, MCM3 protein is readily detected by immunohistochemistry (IHC) in most tissues, particularly liver, lung, thymus, brown adipose tissue and the subventricular zone of the brain. In Mcm3 Lox/Lox embryos, the strong MCM3 staining in the liver was markedly reduced while gH2AX staining became apparent (Fig. 2). Mcm3 mRNA and protein levels confirmed the downregulation of Mcm3 expression in fetal liver extracts ( Supplementary Fig. 4A,B). Global levels of cell proliferation were not affected, as monitored by Ki67 IHC staining ( Supplementary  Fig. 4C).
The liver is the major site of fetal hematopoiesis from E11 until shortly before birth, when this process becomes established in the bone marrow. To evaluate the impact of Mcm3 downregulation in hematopoiesis, specific blood lineages were detected by IHC in mid-late gestation embryos. This approach did not reveal major differences in the abundance of mature B-lymphocytes (stained with Pax-5) and megakaryocytes (stained with FVIII) in the fetal liver, nor T-lymphocytes (stained with CD3) in the fetal thymus of Mcm3 Lox/Lox embryos compared to their Mcm3 þ / þ or Mcm3 þ /Lox counterparts ( Supplementary Fig. 5). In contrast, the staining with erythrocyte marker Ter119 was much reduced in the liver of Mcm3 Lox/Lox embryos (Fig. 3a). Peripheral blood from Mcm3-deficient E14.5 embryos displayed a marked decrease in red blood cells (RBCs), haemoglobin concentration and hematocrit (Fig. 3b), all consistent with a phenotype of fetal anemia. Giemsa staining of peripheral blood cells revealed a large accumulation of immature erythroblasts (EBs) with uncompacted nuclei, as well as the presence of micronuclei that are indicative of genomic instability (Fig. 3c,d).
Impaired erythrocyte maturation upon transplantation. The fetal liver contains HSCs capable of reconstituting the entire hematopoietic system. In order to compare the maturation ability of wild-type and Mcm3-deficient cells towards the erythroid lineage, fetal liver cells (E14.5) derived from wild-type or Mcm3deficient embryos were transplanted into lethally-irradiated recipient mice, in competition (1:1 ratio) with bone marrow cells that constitutively express a Tomato (Tom) fluorescent marker 29 (Fig. 4a). Two months after transplantation, the contribution of control Mcm3 þ / þ cells (Tom-negative) towards mature RBCs in peripheral blood was approximately 50% in the chimaeras (Fig. 4b). In contrast, the contribution of Mcm3-deficient cells (Tom-negative) towards RBCs was much reduced, with Mcm3proficient cells (Tom-positive) taking over most of the RBC population (Fig. 4b). These results confirm that EBs with reduced expression of Mcm3 do not mature properly under steady-state conditions. Chimaeric Mcm3 þ / þ (Tom þ ) / Mcm3 Lox/Lox (Tom-) mice were also used to monitor the acute response to erythropoietic stress induced by phlebotomy (Fig. 4c). In this setting, cells in the spleen with Ter119 high content can be analyzed using the forward scatter parameter (FSC) and CD71 to resolve three EB subpopulations previously labelled as EryA, EryB and EryC 30 . EryA (Ter119 high CD71 high FSC high ) correspond to baso-EBs; EryB (Ter119 high CD71 high FSC low ) correspond to late baso-EBs and polychromatic-EBs; EryC (Ter119 high CD71 low FSC low ) are orth-EBs and reticulocytes. In response to stress, the EryA and EryB subpopulations are rapidly increased 30 . Both Mcm3-competent and Mcm3-deficient EBs reacted to stress, as indicated by the accumulation of splenic EryA and EryB progenitors (Fig. 4d,e). However, the accumulation of EryA in Mcm3-deficient cells was almost 3-fold higher than in Mcm3 þ / þ cells, suggesting that the rapid RBC maturation required under stress conditions is also delayed or impaired (Fig. 4e, right).
Aberrant DNA replication during erythrocyte maturation. To monitor whether inefficient erythrocyte maturation was related to DNA replication and cell cycle progression, fetal liver cells obtained from Mcm3 þ / þ , Mcm3 þ /Lox and Mcm3 Lox/Lox embryos were sorted at the R1-R4 stages (R5 was not included as most cells have started the enucleation process). Interestingly, Mcm3 Lox/Lox baso-EBs (R3) and chr-EBs (R4) displayed a prominent accumulation in S phase, indicative of problems to complete DNA replication ( Fig. 5a). High levels of gH2AX and a higher frequency of fork asymmetry (indicative of stalled forks) were detected in Mcm3 Lox/Lox early EBs, consistent with RS interfering with RBC maturation (Fig. 5b,c).
Next, we analyzed the patterns of origin activity in EB populations sorted from embryonic livers using stretched DNA fibers labelled with thymidine analogs chloro-deoxyuridine (CldU) and iodo-deoxyuridine (IdU). These analyses revealed a dynamic pattern in which more origins are activated as EBs progress in the differentiation pathway. The median IOD was progressively reduced from approximately 100 Kb in pro-EB (R1) to 74 Kb in early baso-EBs (R2), 69 Kb in late baso-EBs (R3) and 56 Kb in orth-EBs (R4; Fig. 5d, left). The length of DNA fibers was similar in all populations, ruling out indirect effects on IOD measurements ( Supplementary Fig. 7). Furthermore, the tendency towards more origin activity was confirmed by measuring origin density (Fig. 5d, right and Table 1). Interestingly, the patterns of origin activity were altered in Mcm3-deficient EBs, which displayed lower IOD values and slightly higher origin density ( Fig. 5d and Table 1). As discussed below, this behaviour likely reflects the increased pressure on early EBs to replicate as a compensatory mechanism for the lack of mature RBCs.
Mcm3 Lox/Lox lethality can be rescued by CHK1 overexpression. Because our results strongly point to RS as the underlying cause for incomplete RBC maturation, we considered the possibility that the embryonic lethality of the Mcm3 Lox/Lox genotype could be alleviated in a background with higher tolerance to RS, such as the recently described strain carrying an extra copy of checkpoint kinase CHK1 (ref. and multipotent progenitor cells (MPPs: Lin À /cKit þ /Sca1 þ /Flk2 þ ), had lower amounts of MCM3 protein and higher levels of RS marker gH2AX (Fig. 6a). Intriguingly, single-cell tracking of cell division revealed slightly faster proliferation kinetics of Mcm3 Lox/Lox than Mcm3 þ / þ fetal HSCs (Fig. 6b). This could reflect a pressure to proliferate and differentiate into MPPs, as it has been described for adult HSCs in the presence of DNA damage 33 . Consisting with this notion, immunophenotyping analyses revealed a two-fold increase in the concentration of LSK cells in Mcm3 Lox/Lox embryos, caused by the accumulation of MPPs (Fig. 6c).
To directly test their functionality, fetal HSCs were isolated from E14.5-E16.5 CD45.2 C57BL/6 Mcm3 þ / þ , Mcm3 þ /Lox and Mcm3 Lox/Lox embryos and transplanted into lethally irradiated congenic CD45.1 C57BL/6 recipient mice. Four months after transplantation, recipient mice displayed decreased Mcm3 Lox/Lox chimerism in all hematopoietic organs (Fig. 6d). Because formation of an adult HSC compartment was observed (Fig. 6d,  right), donor-derived HSCs were re-isolated from the bone marrow of recipient mice to evaluate their proliferation potential in vitro and in vivo. Mcm3-deficient adult HSCs displayed RS markers gH2AX and RPA (Fig. 6e) and inefficient colony formation in methylcellulose in the presence of aphidicolin ( Fig. 6f). Furthermore, when subjected to a major replication challenge such as a secondary transplantation, Mcm3 Lox/Lox adult HSCs were drastically affected in their ability to self-renew and regenerate the blood system (Fig. 6g).

Discussion
Mcm2-7 are essential genes that encode key components of the main DNA helicase involved in genome replication. Mcm2-7 are highly expressed in proliferating cells, and DNA replication can still occur in cell lines after a significant reduction (490% in some cases) in the cellular concentration of MCM protein complexes [17][18][19] . This tolerance to MCM downregulation is related to the fact that MCM complexes are engaged with DNA in large excess relative to the number of replication forks normally established during S phase. The surplus of MCM complexes license dormant origins that are activated only as a rescue mechanism when DNA replication is disrupted (reviewed in reference 34 ). The use of dormant origins in vivo is well documented in the mouse, and MCM downregulation beyond B2/3 of its physiological levels causes embryonic lethality or promotes tumorigenesis in adults [7][8][9][10][11][12][13] . Therefore, it is likely that certain cell types in the developing embryo, such as stem and progenitor cells, are particularly sensitive to RS induced by low MCM levels.
We have tested the hypothesis that different cell types may have different requirements for MCM concentration using a novel strain with hypomorphic Mcm3 expression. While Mcm3 Lox/Lox MEFs proliferated and replicated DNA with approximately 1/3 of the normal concentration of MCM3 protein, a similar reduction severely impaired hematopoietic progenitors, indicating a stricter requirement for origin licensing in the latter. In mid-gestation, hematopoiesis in the fetal liver is largely geared towards the production of RBCs to guarantee oxygen delivery to the rapidly growing embryo. Interestingly, erythroid precursors undergo several rounds of DNA replication and cell division during terminal differentiation, and genetic models that ablate cell cycle regulators such as Rb, E2F4, E2F8 or D-cyclins frequently result in embryonic anemia [35][36][37][38] . While D-Cyclin/CDK and Rb/E2F constitute the axis of a large transcriptional pathway regulating multiple genes, here we report for the first time that downregulation of a single MCM gene is sufficient to impair hematopoietic progenitor cells, causing anemia. Cytological analyses of embryonic blood revealed lower counts of RBCs and abundance of immature nucleated erythroblasts.   approximately 1/3 of its normal concentration, the severity of the anemia causes embryonic lethality in the C57BL/6 genetic background. As reported in other mouse models such as the Mcm4-chaos mutant 8,11,12 or the Rif1 KO mouse 39 , Mcm3 Lox/Lox embryonic lethality was partially alleviated in a mixed C57BL/6-CD1 background and it was further rescued by overexpression of CHK1 kinase, reinforcing the connection between RS and the phenotypes observed.
To our knowledge, the single-molecule analyses of DNA replication in EB precursors isolated from the fetal liver provide the first evidence that the program of DNA replication undergoes active changes during the physiological maturation of mammalian erythrocytes. As pro-EBs proliferate and differentiate into mature reticulocytes, their replication program requires the progressive activation of more origins. This observation has interesting antecedents: chicken erythrocytic progenitors forced    to differentiate ex vivo display a broadening in origin usage in the b-globin locus 23 . Also, origin firing is enhanced in murine erythroleukemia cells derived from transformed pro-EBs and forced to differentiate in culture 40 . It is conceivable that rapid DNA replication, driven from an increasingly high number of origins, is related to the rapid loss of DNA methylation marks observed during mouse erythropoiesis 41 . These patterns of DNA replication were altered in Mcm3-deficient pro-EBs, which displayed a higher frequency of origin activation and asymmetric forks from the start of their differentiation program. Later in differentiation, baso-EBs and chr-EBs were not capable of completing DNA replication, accumulated in S phase and activated the apoptotic program, thus preventing erythrocyte maturation. This likely triggers a compensation mechanism that pressures pro-EBs to replicate and proliferate, contributing to the aberrant replication patterns. reflects improper coordination between fork pairs 42 and may also trigger the activation of compensatory origins. The fact that pro-EBs were capable of activating extra origins after Mcm3 downregulation underscores the fact that a very large excess of MCM proteins associate with DNA during G1. Mid gestation is also the time of HSCs expansion in the fetal liver. At this stage, HSCs divide very rapidly (reviewed in reference 43 ), which could also make them vulnerable to RS. The capacity of Mcm3 Lox/Lox fetal HSCs to reconstitute the immune system upon transplantation into irradiated recipients was compromised but not completely impaired, and a population of donor-derived adult HSCs was generated in recipient mice. The lack of engraftment potential was strikingly exacerbated when Mcm3 Lox/Lox HSCs were re-isolated from the bone marrow of recipient mice and tested in secondary transplantations. These observations are in line with the emerging concept that RS affects HSC functionality in older individuals. With age, HSCs lose reconstitution potential for causes that have been debated [44][45][46][47][48] (reviewed in references 49,50 ). Interestingly, a comparison of gene expression profiles of young versus old adult HSCs revealed that Mcm2-7 are downregulated in old HSCs, suggesting a new mechanistic link between faulty DNA replication and functional impairment 32 . The work presented here with a new mouse model with hypomorphic MCM expression provides direct genetic evidence that RS is a driving force behind HSC loss of functionality.
In summary, Mcm3 downregulation during embryonic development caused RS in hematopoietic progenitors leading to fetal anemia; in adults, it reduced life expectancy and promoted lymphomagenesis. Thus, future therapies designed to modulate RS are a promising way to fight ageing and hematopoietic malignancies.

Methods
Generation of Mcm3 þ /Lox mouse strain. A targeting vector was designed that included an IRES-EGFP-luciferase reporter cassette after the Mcm3 stop codon located at exon 17, two loxP sites flanking Mcm3 exons 14 to 17, and a FRTflanked neomycin-resistance cassette for selection. The linearised vector was electroporated into 129Sv/Pas ES cells, and its genomic integration by recombination was screened by PCR and Southern blotting. Verified transgenic ES clones (Mcm3-Lox-Neo) were microinjected into C57BL/6J blastocysts and germline transmission was confirmed. Vector preparation and genetic manipulations were conducted at Genoway (Lyon, France). Mice carrying the Mcm3-Lox-Neo allele were crossbred with a strain expressing Flp recombinase to eliminate the neomycin-resistance cassette, generating the Mcm3-Lox allele. Mcm3 þ /Lox mice were crossbred with the CMV-Cre strain to generate the Mcm3-null allele, referred to as Mcm3 À . Mice of both sexes were used in experiments, except when noted otherwise. Primers used for PCR genotyping are indicated in Supplementary  Table 4.
Mouse handling and luciferase detection. Mice were hosted at the CNIO animal facility, except for a cohort of Mcm3 þ /Lox mice that were transferred to the UCSF animal facility for the analyses of HSCs. Animal procedures were approved by the Ethical Committee of the Instituto de Salud Carlos III (Madrid, Spain), and HSC transplantation experiments were performed under UCSF IACUC-approved protocols. For each experiment, sample size was calculated using Resource Equation 51 . In vivo luciferase imaging was performed in an IVIS imaging system (Caliper Life Sciences, CA) in mice anesthetized with 2.5% isoflurane, 10 min after an intraperitoneal injection of luciferin (150 mg kg À 1 of body weight). Mice were partially shaved to monitor bioluminescence in the back skin. When indicated, mice were sacrificed 10 min after luciferin injection and necropsies were performed in order to monitor luciferase activity in internal organs.
Mouse embryonic fibroblasts (MEFs) isolation and culture. Mcm3 þ / þ , Mcm3 þ /Lox and Mcm3 Lox/Lox primary MEFs were derived from E12.5-14.5 embryos resulting from crosses between Mcm3 þ /Lox mice and cultured in Dulbecco's modified minimal Eagle medium (DMEM) supplemented with 10% FBS and antibiotics. For cell proliferation curves, aliquots of 0.5 Â 10 5 cells were seeded and counted every 2 days in a hemocytometer. When indicated, 10 mM bromo-deoxyuridine (BrdU; Sigma) was added to the medium for 30 min before cell harvesting. Cells were fixed in 70% ethanol, washed in PBS and stained with 50 mg ml À 1 propidium iodide (Sigma) in the presence of 10 mg ml À 1 RNase A (Qiagen). Fixed cells were treated with 2 M HCl for 20 min and incubated with FITC-conjugated anti-BrdU antibody for 60 min. Flow cytometry data was acquired in a FACS Canto II (BD, San Jose, CA) and analyzed with FlowJo 9.7.5 (Tree Star, Ashland, OR).
Quantitative RT-PCR, cell extracts and immunoblots. Total RNA was extracted using the RNeasy Mini Kit (Qiagen). 1 mg of total RNA was used for randompriming cDNA synthesis with SuperScript II (Invitrogen), and quantitative PCR was performed with Power SYBR Green master mix in an Applied Biosystems 7900HT Fast qRT-PCR machine. Primers used for gene expression are indicated in Supplementary Table 5. Extracts were prepared by direct suspension of cells in Laemmli buffer followed by three pulses of sonication for 15 s at 15% amplitude (Branson Digital Sonifier). SDS-PAGE and immunoblots were performed using standard methods.
Mcm3-Lox mRNA sequencing. Total RNA from Mcm3 Lox/Lox MEFs was isolated and cDNA was prepared as indicated in the previous section. Three overlapping fragments covering the whole Mcm3-Lox cDNA molecule were amplified by PCR (primer sequences are shown in Supplementary Table 6) and sequenced. Experimental sequences were compared with the targeting vector used by Genoway and the NCBI reference sequence for Mcm3 mRNA (NM_008563.2). DNA and encoded protein alignments were performed with Clustal Omega (EMBL-EBI website).
Immunofluorescence microscopy. Primary MEFs or primary erythroid populations were incubated in mCLEAR bottom polylysine-treated 96-well or 384-well plates (Greiner Bio-One) for 2 h at 37°C, fixed in 4% paraformaldehyde in PBS for 15 min at RT and permeabilizad with 0.5% Triton-X100 in PBS (5 min at RT). Cells were incubated in blocking solution (1% bovine serum albumin in PBS) for 30 min. Primary antibody solution was applied for 1 h at RT. For RPA immunostaining, soluble proteins were extracted prior to fixation with 0.5% Triton X-100 in CSK buffer (10 mM Pipes-KOH pH 7.0, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 ). Cell nuclei were stained with DAPI (Sigma). Images were acquired either in a Leica-TCS SP5X confocal microscope, with a HCX PL APO 20x objective using LAS AF software or in an Opera High-Content Screening System (PerkinElmer) with an APO 20 Â , 0.7 NA water-immersion objective using Acapella software (PerkinElmer). To estimate the number of RPA foci-positive cells or gH2AX focipositive cells, 100-200 or 250-500 cells were scored respectively in each condition. To measure intensity of MCM3 or gH2AX in erythroid populations, images were acquired from each well, nuclei were marked by DAPI staining and protein immunostaining intensity was measured within the nuclei.
Single-molecule analysis of DNA replication. Exponentially growing MEFs or single-cell suspensions of fetal livers were pulse-labeled with 50 mM CldU (20 min) followed by 250 mM IdU (20 min). Labeled cells were harvested and resuspended in 0.2 M Tris pH 7.4, 50 mM EDTA and 0.5% SDS. Stretched DNA fibers were prepared as described 52 . For immunodetection of labeled tracks, fibers were incubated with primary antibodies for 1 h at RT and the corresponding secondary antibodies for 30 min at RT, in a humidity chamber. DNA was stained with anti-ssDNA to assess fiber integrity. Fiber images were obtained in a DM6000 B Leica microscope with an HCX PL APO 40 Â , 0.75 NA objective. The conversion factor used was 1 mm ¼ 2.59 kb. In each assay, 4300 individual tracks were measured for FR estimation, 4100 fibers containing two or more origins were analyzed for IOD estimation, 4200 labeled fibers were counted for fiber length and origin density, and 4180 bidirectional forks were counted for fork symmetry.
Steady-state and stress erythropoiesis response. Mice with constitutive expression of fluorescent tdTomato from the Rosa26 locus were generated by germline recombination of the Rosa-CAG-LSL-tdTomato allele 29 , carried by the Ai14 reporter mouse line (stock 007914, Jackson Laboratory). Total cells were recovered from the bone marrow of CAG-Tomato mice and from the fetal livers of E14.5 Mcm3 þ / þ or Mcm3 Lox/Lox C57Bl/6 donor mice, and transplanted into lethally irradiated C57BL/6 primary female recipients (10 6 cells per mouse; 5 mice per group). 8 weeks post-transplantation, donor chimerism (Tom þ /Tom-) was analyzed in peripheral blood after staining with Ter119. To monitor the erythropoietic stress response, mice were subjected to phleblotomy (400 ml peripheral blood per 25 g of body weight). 4 days post-phlebotomy, accumulation of RBC progenitors EryA, EryB and EryC in the spleen was monitored as described 30 , differentiating between Tom þ and Tom-cells.
Single-cell tracking of division kinetics. HSCs isolated from the fetal liver of E14.5-16.5 embryos were sorted into 96-well plates (1 cell/ well) and visually inspected after 12 h to confirm successful single cell sort. At the indicated time points, cell cultures were monitored to establish the kinetics of the first cell division (appearance of 2 or more cells). 96 wells were scored per condition.
Statistical methods. In column graphs, data are expressed as mean±s.d. Statistical analyses were done using Fisheŕs test. When the data are presented in scatter dot plots, the bar corresponds to the median value. For the analyses of FR, IOD and fork symmetry parameters in stretched DNA fibers and immunofluorescence data from acquired in the high-throughput Opera system, data distribution was normally not Gaussian, and differences between samples were assessed with the nonparametric Mann-Whitney rank-sum test. In the analysis of apoptosis and cell death in EB populations, the mean values derived from 3 independent experiments were compared by One-Way Anova and further Bonferroni post-test. Statistical analysis was performed in Prism v4.0 (GraphPad Software).