Bb1-3, a transgenic hybrid cell line with erythroid and megakaryocytic differentiation potential that expresses high levels of human γ-globin and human β-globin


Dr Thomson MRC Molecular Haematology Unit, Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DS.


We have characterized a murine hybrid cell line, Bb1-3, generated by the fusion of mouse primary erythroblasts with MEL cells. It proliferated in serum-free medium and displayed a low level of spontaneous erythroid and megakaryocyte differentiation. Terminal erythroid differentiation could be induced with HMBA and DMSO and was enhanced by serum. Treatment with phorbol esters resulted in a high proportion of megakaryocytes and the expression of megakaryocytic specific lineage markers. Bb1-3 cells contain a human β-globin transgene that was expressed at levels of 20–50% of the endogenous mouse globin genes. Initially, expression was largely limited to the β-globin gene but after adaptation to serum free growth, equal expression of both the human γ- and human β-globin genes was observed. This cell line provides further evidence that the differentiation potential of mouse erythroleukaemia cells is not restricted to the erythroid lineage and should be useful to study the mechanisms underlying both developmental globin gene regulation and the terminal differentiation of bipotential erythroid/megakaryocytic progenitor cells.

Immortalized haemopoietic cell lines have been used extensively to study the regulation of lineage-restricted gene expression. In particular, mouse erythroleukaemia (MEL) cells have provided a valuable model system to analyse the mechanisms involved in regulating the globin gene loci. MEL cells are generated following the infection of susceptible mice with the Friend virus complex, which is composed of two distinct retroviral entities: a replication-defective spleen focus forming virus (SFFV) and a replication competent Friend murine leukaemia virus (F-MuLV) ( Friend, 1957; Ben David & Bernstein, 1991). The immortalized cells obtained following Friend virus immortalization have been considered to represent committed unipotential erythroid progenitor cells ( Stamatoyannopoulos & Nienhuis, 1994). Recently, however, evidence has been presented suggesting that MEL cells also express phenotypic markers from the megakaryocyte lineage, including acetylcholine esterase, platelet factor 4 (PF4), von Willebrand factor and glycoprotein IIb ( Paoletti et al, 1995 ; Vannucchi et al, 1997 ) as well as the megakaryocyte-specific antigen recognized by the monoclonal antibody 4A5 ( Paoletti et al, 1995 ; Burstein et al, 1992 ).

We have characterized a murine hybrid cell line based on MEL cells, that may be induced along the erythroid or megakaryocytic pathways, which is also capable of expressing high levels of human γ- and β-globin mRNA. These cells may prove useful for studies concerning both haemopoietic lineage commitment and developmental globin gene regulation.


Generation of the cell lines Bb1 and Bb1-3

The parental cell line was generated by the fusion of murine erythroblasts transgenic for a 40 kb human β-globin construct (HS2 GγAγ−117δβ) ( Morley et al, 1991 ) with mouse erythroleukaemia (MEL) cells using standard techniques ( Stanworth et al, 1995 ). The Bb1 cell line was obtained from the initial hybrid pool by limited dilution cloning and maintained in Iscove's modified Dulbecco's medium plus 17.5% fetal calf serum, 2 m ML-glutamine, 25 units/ml penicillin and 25 μg/ml streptomycin.

Bb1 was adapted to grow in serum-free medium by the gradual reduction of fetal calf serum in the medium to a concentration of 0.5% ( Harlow & Lane, 1988), before transfer to a serum-free environment (modified from Heyworth & Spooncer, 1993): Iscove's modified Dulbecco's medium (IMDM); 2 m ML-glutamine; 25 units/ml penicillin; 25 μg/ml streptomycin; phospholipid emulsion (25 μg/ml Intralipid, Kabi); cholesterol (7.8 μg/ml, Sigma); linoleic acid (5.6 μg/ml, Sigma); bovine serum albumin (10 mg/ml, Boehringer Mannheim); bovine holotransferrin (98% saturated, 300 μg/ml, Sigma); bovine insulin (6 μg/ml, CP Pharmaceuticals); α-monothioglycerol (100 μM, Sigma) and sodium pyruvate (1 m M, Gibco BRL). After a further 5 months, Bb1 was re-cloned in serum-free methylcellulose to generate the cell lines Bb1-1 to Bb1-6, all of which had morphological and growth features similar to the parental cell line. The Bb1-3 cell line was selected for further analysis.

Inducers and growth factors

The effects on cell growth and differentiation of the following inducers of differentiation were assessed on Bb1-3 in serum-free medium for periods of up to 7 d: 20% fetal calf serum; 3 m M hexamethylene-bis-acetamide (HMBA) (Sigma); 1.5% dimethylsulphoxide (DMSO) (Sigma); 25 m M haemin (Boehringer Mannheim); 10−7 M or 10−8 M phorbol 12-myristate 13-acetate (TPA), 10−6–10−9 M all-trans retinoic acid (Sigma).

Cytokines were added to Bb1-3 in serum free medium at the following concentrations: murine IL-3 (Sigma) 50 units/ml; murine IL-6 (Sigma) 20 ng/ml; murine GM-CSF (Sigma) 200 units/ml; human G-CSF (Boehringer Mannheim) 100 units/ml; murine M-CSF (Sigma) 50 units/ml; and murine erythropoietin (Boehringer Mannheim) 2 units/ml.

Haemoglobin quantitation

The effect of inducers of erythroid differentiation was assessed by staining the cells using a protocol adapted from Orkin et al (1975 ). The cellular haemoglobin content was measured by a modified benzidine-peroxide assay as described by Clarke et al (1982 ).

Acetylcholine esterase staining

To stain for the murine megakaryocytic form of acetylcholine esterase, air-dried cytospin slides were treated according to the method of Testa & Molineux (1993).

DNA analysis

Southern blot analysis was performed using standard methods ( Southern, 1975; Sambrook et al, 1989 ) with the following probes: Pst I-Pst I murine spi-1 probe A ( Moreau-Gachelin et al, 1988 ); Hind III-Bgl II mouse p53 cDNA ( Jenkins et al, 1984 ); EcoR I-EcoR I mouse fli-1 B1 and EcoR I-EcoR I mouse fli-1 B2 probes ( Ben-David et al, 1990); mouse p45 NF-E2 cDNA ( Andrews et al, 1993 ); mouse c-mpl probe B ( Skoda et al, 1993 ); Bam HI-Pvu II E57BS probe ( Penciolelli et al, 1987 ); Bgl II-Pvu II human β-globin LCR hypersensitive site 2 (unpublished) and the Bam HI-EcoR I IVS2β fragment from the human β-globin gene ( Morley et al, 1992 ).

In addition, a probe specific for the Friend murine leukaemia virus LTR was synthesized by PCR using primers designed from the Friend murine leukaemia virus (Clone 57) U3 region (accession number X02794): F1: 5′ GAA CAG ATA CGC TGG GCC AAA C (co-ordinates 8026–8047); and R1: 5′ GAC CGG GCC GAA ACT GCT TAC C (co-ordinates 8080–8059).

DNase I hypersensitive site analysis

DNase I hypersensitive assays were performed by treating nuclei with 0–160 units of DNase I (Pharmacia) according to the method of Higgs et al (1990 ).

RNA analysis

RNA was extracted from cell lines by the method of Chomczynski & Sacchi (1987) and gene expression was assessed by RNase protection assay ( Zinn et al, 1983 ). The probes used were: mouse beta h1 globin (pT7/T3-18βh1) ( Stanworth et al, 1995 ); mouse alpha globin (pSPJMαS) ( Higgs et al, 1990 ); mouse zeta globin (pSP64Mζ), mouse beta major globin (pSP64Mβ), mouse epsilon globin (pSP64Mɛ) and human alpha globin (pSP64α132) (all from Baron & Maniatis, 1986); human zeta globin (pζ2SP64, Vyas et al, 1992 ); human beta globin (pGβ1) and human gamma globin (pGGγ2.1) (both Morley et al, 1991 ); rat beta actin (pAM 19) ( Nudel et al, 1983 ).

A probe for mouse platelet factor 4 was generated by PCR using primers F1: 5′ TGG TTG CTG TCA CCA GGG (co-ordinates 553–570) and R1: 5′ CTG GGG AAC CGC ACA GTG G (co-ordinates 953–934) from the rat PF4 gene (accession number M15254); this was cloned into pCRScript (Stratagene), linearized with Hind III, and transcribed with T3 polymerase to give a 483 base riboprobe which generates a protected fragment of 132 bp.

Analysis of GPIIb expression was undertaken using an EcoR I-Rsa I fragment covering exon 1 of the mouse GPIIb gene that was cloned into pSPT 18 (Boehringer Mannheim) and linearized with Mlu NI. A 324 base antisense transcript was synthesized with Sp6 polymerase that protects a fragment of 185 base pairs.

Quantitation of RNA was performed either by scintillation counting, after excising the appropriate protected bands from the gel, or by phosphoimager analysis, and then correcting for the number of G residues in the protected fragment.

Karyotype analysis

Chromosome numbers were determined by analysis of metaphase spreads prepared by the method of Buckle & Rack (1993).


Among the hybrids emerging from a fusion between transgenic mouse erythroblasts and MEL cells, one line was unusual in that in addition to the normal blast-like appearance of the majority of cells, spontaneously differentiated cells of both the erythroid and megakaryocyte lineages were present in appreciable numbers (Fig 1A). All other hybrid lines that we have examined were morphologically similar to the parental MEL cells, did not show spontaneous erythroid differentiation and did not contain cells with the morphology of megakaryocytes.

Figure 1.

0−7 M TPA.

The cells were cloned; all 10 clones examined retained the same phenotype, suggesting they were possibly derived from a single hybrid cell. One of these clones, Bb1, was adapted for growth in serum-free conditions. Its phenotype remained unaltered following this adaptation, and through further recloning in serum-free methylcellulose to generate the subline Bb1-3. The Bb1-3 line has remained morphologically stable for more than 2 years in continuous culture.

DNA analysis

As several other haemopoietic cell lines were being derived in the laboratory at the same time, including cells immortalized by the murine retrovirus MPLV, it was important to demonstrate that this unusual line was hybrid in origin and not due to cellular contamination. The c-mpl probe B ( Skoda et al, 1993 ), which also covers the majority of the transduced oncogene in MPLV, detected only the endogenous c-mpl gene, with no additional bands indicative of the presence of MPLV proviral genomes in Bb1 and Bb1-3 (data not shown).

The genomes of mouse erythroleukaemia cell lines are often associated with the disruption of the ets genes, spi-1 (PU.1) ( Moreau-Gachelin et al, 1988 ; Klemsz et al, 1990 ) and fli-1 ( Ben-David et al, 1990 , 1991; Sels et al, 1992 ) as well as commonly showing mutations of the tumour suppressor gene, p53 ( Mowat et al, 1985 ; Rovinski et al, 1987 ; Hicks & Mowat, 1988; Ben-David & Bernstein, 1991; Lu et al, 1994 ) or the fli-2 locus ( Lu et al, 1994 ). No rearrangements were noted in either the fli-1 or fli-2 loci in Bb1 and Bb1-3, but gross rearrangements were obvious at the p53 (Fig 2A) and spi-1 (Fig 2B) loci. These rearrangements were identical in Bb1 and Bb1-3 cells and the MEL cell line used to generate them.

Figure 2.

β probes, showing the identical pattern in Bb1-3 and the J14 transgenic mouse line from which the hybrid was derived.

Contribution to the Bb1-3 genome by MEL was confirmed by mapping the integration pattern of F-MuLV proviral sequences in Bb1-3 and MEL. A probe was generated that specifically hybridized to the U3 region of both the 5′ and 3′ LTR of F-MuLV, but not to the LTR sequences of endogenous proviral genomes, owing to a 10 bp deletion in F-MuLV relative to all of the other murine retroviruses in the database. Using this probe, the same pattern of approximately seven F-MuLV proviral genomes was observed in Bb1-3 and in MEL (data not shown); this observation was confirmed using the E57 BS probe of Penciolelli et al (1987 ) (Fig 2C). The presence, therefore, of F-MuLV sequences with an identical pattern of integration along with identical rearrangements of the p53 and spi-1 loci indicated that MEL contributed substantially to the genome of Bb1 and Bb1-3.

As well as containing genomic sequences from a MEL cell line, Bb1-3 contains the human beta globin transgene HS2 GγAγ−117δβ from the transgenic mouse line J14. As may be seen from Fig 2D, the pattern of internal bands, and 5′ and 3′ junction fragments, in Bb1-3 DNA was identical to those of the J14 transgenic mouse line and confirmed the hybrid nature of this cell line.

We also checked the chromosome composition of Bb1-3. When first examined, the modal chromosome number of Bb1-3 was 53, with a range of 50–61. This compares with 40 chromosomes in mouse erythroblasts and a modal number of 39 (range 37–42) in the MEL cells and is consistent with the karyotype of a hybrid cell line. One year later the modal number in the Bb1-3 cells remained at 53, indicating a stable karyotype (data not shown).

Differentiation potential of Bb1-3

Various combinations of haemopoietic growth factors (IL-6 + IL-3 + GM-CSF + G-CSF; IL-6 + IL-3 + GM-CSF + M-CSF; IL-6 + IL-3 + GM-CSF + Epo; IL-6 + GM-CSF + M-CSF; IL-6 + IL-3 + Epo; GM-CSF + G-CSF) had no effect on the pattern of cell growth or differentiation when added to Bb1-3 in serum-free medium. In particular, no evidence for granulocytic, mast or macrophage differentiation was observed in these cells, nor did the addition of mouse recombinant erythropoietin significantly enhance erythroid differentiation. No effect was observed after treatment of the cells with all-trans retinoic acid.

Erythroid differentiation of Bb1-3

When Bb1-3 cells, grown in serum-free medium, were transferred back to IMDM containing 20% fetal calf serum, erythroid differentiation markedly increased. Several chemical inducers of erythroid differentiation were then tested for their effect on haemoglobin production by Bb1-3 in the presence or absence of serum ( Table I). Although megakaryocytes remained, the addition of either HMBA or DMSO resulted in marked erythroid differentiation, with the majority of cells developing an erythroid appearance after 3 d and some maturing to late normoblasts after 7 d; a further enhancement of erythroid differentiation was obtained upon the addition of haemin (Fig 1B; Table I). These results were borne out by mRNA analysis, where a >10-fold increase in mouse globin mRNA was observed and a combination of 20% fetal calf serum, 3 m M HMBA and 25 μM haemin was found to be optimal to induce globin gene expression in Bb1-3, with up to a 50-fold increase in globin RNA levels (Fig 3A). A decrease in globin mRNA was observed following treatment with TPA (Fig 3A).

Table 1. Table I. The effect of fetal calf serum, erythropoietin, and various chemical inducers on globin gene expression in the Bb1-3 cell line.Thumbnail image of
Figure 3.

hybridized with mouse α-globin and rat β-actin probes to show the induction of globin mRNA under various conditions. (B) RNase protection assay with human γ-globin, human β-globin and mouse α-globin probes to demonstrate the changing pattern of globin gene expression upon adaptation of Bb1 to growth in serum-free medium.

Erythroid induction was also monitored by measuring the increase in the amount of haemoglobin synthesized by the Bb1-3 cell line. Uninduced Bb1-3 contain on average 0.36 pg haemoglobin per cell, which is slightly higher than uninduced MEL cells. After 3 d induction under optimal conditions the mean level of haemoglobin per cell increased approximately 54-fold, achieving an average of 19.5 pg of haemoglobin per cell ( Table II). Benzidine staining of cytospin samples showed that >90% of cells were positive for haem and that there was marked intercellular variability in the haemoglobin content within each cell, with the most densely staining cells having late normoblast morphology (Fig 1C).

Table 2. Table II. The effect of serum, haemin and HMBA on haemoglobin production in MEL and the Bb1-3 cell lines. Human lymphoblastoid cells were assayed to correct for the presence of haemin on the measurement of haemoglobin concentration.Thumbnail image of

Megakaryocytic differentiation of Bb1-3

Low numbers of cells with the morphology of megakaryocytes persisted through two separate cloning procedures. Staining for the megakaryocytic form of acetylcholine esterase was positive in 0.1% of the Bb1-3 cells compared to <0.01% of the parental MEL cell line. In order to determine whether the megakaryocytic component of Bb1-3 could be further induced, the cells were grown in serum-free medium in the presence of phorbol 12-myristate 13-acetate (TPA). With 10−7 M TPA, no obvious change in Bb1-3 cell morphology was observed after 3 d; however, 30% of Bb1-3 cells had the morphological appearance of mature megakaryocytes (Fig 1D) and 11% of all Bb1-3 cells were strongly acetylcholine esterase positive by day 7 (Fig 1E). Similar treatment of MEL cells resulted in no more than 0.1% of the cells being acetylcholine esterase positive (Fig 1F). Despite the appearance of megakaryocytes after the treatment of Bb1-3 with 10−7 M TPA for 7 d, an analysis of propidium iodide stained nuclei for DNA content did not provide evidence for endoreduplication, suggesting that these cells were multinucleate rather than having a single multilobulated nucleus (data not shown).

The capacity of Bb1-3 for megakaryocyte gene expression was analysed by RNase protection using probes to mouse platelet factor 4 and glycoprotein IIb. As may be seen in Fig 4, platelet factor 4 mRNA and GPIIb were readily detectable in Bb1-3 cells prior to induction. Upon treatment with TPA there were marked increases in the level of both PF4 and GPIIb mRNA ( Table III); a fall in platelet factor 4 was observed after the induction of erythroid differentiation with HMBA and haemin in Bb1-3, but not in MEL, consistent with the observations of Vannucchi et al (1997 ).

Figure 4.

mRNA decreases in Bb1-3 following induction with either HMBA and haemin, but increases on treatment for 7 d with 10−7 M TPA or 10−8 M TPA, (B) increased expression of mouse GPIIb in Bb1-3, and decreased expression in MEL, following induction with 10−7 M TPA.

Table 3. Table III. The effect of TPA on the expression of platelet factor 4 and GPIIb in Bb1-3 and MEL as a percentage of β-actin expression. Thumbnail image of

Globin gene expression in Bb1 and Bb1-3

Analysis of mouse globin gene expression in the Bb1 and Bb1-3 cell lines by the RNase protection assay showed highly inducible levels of α-globin (and βmaj-globin) mRNA (Fig 3A). No endogenous embryonic globin mRNA was detected at any stage (data not shown). The human globin gene expression profile in Bb1, grown in medium with serum, showed levels of human γ/γ + β of almost 7%, whereas the combined level of human γ + β-globin mRNAs were approximately equal to that of mouse α-globin mRNA. The γ/γ + β ratio was unaltered by the addition of any chemical inducers or growth factors, and a similar low level of human γ-globin mRNA was observed in the nine other clones obtained from the initial cloning that produced Bb1.

Following the adaptation of Bb1 to serum-free growth, and in subsequent experiments with various inducers and different batches of fetal calf serum, the proportion of total human globin mRNA/mouse α-globin mRNA averaged 49.5 ± 15.8%. However, in serum-free conditions, the γ/γ + β mRNA ratio increased to 48.2 ± 8.3% and remained at this level in all further studies, irrespective of the presence of growth factors or chemical inducers (Fig 3B). The human γ-globin in these cells comprised approximately equal amounts of Gγ and Aγ mRNA (Gγ/Gγ+Aγ = 53.5 ± 13.5%).

To confirm that the change in growth conditions was responsible for the alteration in the globin gene expression profile, a second aliquot of the original Bb1 cells was recovered from cryostorage and adapted to growth in serum-free medium over a period of approximately 4 weeks. Again, equal amounts of human γ- and β-globin mRNA were observed (data not shown).

DNase I hypersensitivity of the human globin transgene

The transgene in the Bb1-3 hybrid line is under the regulation of the HS2 element from the human β-globin LCR. To determine whether this DNase I hypersensitive site was correctly reformed in this environment, nuclei from uninduced Bb1-3 cells, and cells induced for erythroid maturation, were treated with increasing amounts of DNase I. As shown in Fig 5, the hypersensitive site was readily detected prior to induction of the cells, and remained sensitive following erythroid induction.

Figure 5.

Fig 5. DNase I hypersensitivity of HS 2 of the human β-globin LCR in Bb1-3 nuclei prior to (left) and following (right) erythroid induction with HMBA and haemin. The triangles above the lanes indicate the increasing amounts of DNase I (0–160 u/ml) added to each sample. The DNA was digested with Sac I and hybridized with the Bgl II-Pvu II HS2 probe. The sub-band produced by digestion with DNase I is delineated by an asterix.


Bb1 and Bb1-3 are hybrid cell lines that contain genomic sequences from MEL cells and a transgenic mouse strain harbouring a human β-globin gene construct. Bb1-3 proliferates rapidly in serum-free medium and demonstrates a low level of spontaneous erythroid and megakaryocyte differentiation in the absence of added growth factors. On exposure to the phorbol ester, TPA, Bb1-3 displayed a marked increase in the level of three megakaryocytic markers: acetylcholine esterase, PF4 and GPIIb. Not all of these effects were observed with the MEL cell partner used for hybrid formation. Treatment of the cells with erythroid inducers resulted in their differentiation to late normoblasts and the accumulation of up to 20 pg haemoglobin per cell, approximately twice that achieved in induced parental MEL cells. We have not observed this bilineage phenotype in any of our other hybrid cell lines; presumably the particular chromosomal constitution of this line, with any attendant imbalance in gene expression, was responsible for the unusual phenotype.

Although mouse erythroleukaemia cell lines have previously been considered to be erythroid, and their response to various chemical inducers make them an excellent model for erythroid maturation, there is increasing evidence that megakaryocyte markers are expressed in MEL cells ( Burstein et al, 1992 ; Paoletti et al, 1995 ; Vannucchi et al, 1997 ). Both erythrocytes and megakaryocytes are thought to be derived from a bipotential erythroid/megakaryocyte progenitor cell ( McLeod et al, 1976 ; Suda et al, 1984 ; Nishi et al, 1990 ; Debili et al, 1996 ). Therefore it may be possible that the phenotype of MEL cells reflects the identity of the target cell initially exposed to Friend virus complex rather than a consequence of their immortalization. As MEL cells retain some vestige of their bipotential history they may more easily generate hybrids such as Bb1 and Bb1-3 that can be induced down either pathway and hence should be very useful for looking at the regulation of gene expression in erythroid/megakaryocytic differentiation.

Several human leukaemia cell lines have been described, e.g. MEG-01 ( Ogura et al, 1985 ), JK-1 ( Tani et al, 1996 ), HEL ( Papayannopoulou et al, 1987 ), K562 ( Andersson et al, 1979 ; Rutherford et al, 1981 ), TF-1 ( Kitamura et al, 1989 ), LAMA-84 ( Seigneurin et al, 1987 ), OCIM2 ( Papayannopoulou et al, 1988 ), F-36 ( Chiba et al, 1991 ), UT-7 ( Komatsu et al, 1991 ; Nicolis et al, 1993 ; Hermine et al, 1992 ), UT-7/GM ( Komatsu et al, 1997 ) and MB-02 ( Morgan et al, 1991 ; Broudy et al, 1993 ; Perrine et al, 1989 ), that, like Bb1-3, express both erythroid and megakaryocytic features. Many of these lines have either an absolute requirement for exogenous growth factors or do not display the degree of differentiation upon induction that characterizes Bb1-3. Furthermore, in the main, the levels of globin transcription in these human leukaemia lines are extremely low; chemical induction is usually associated with only a small increase in globin transcription and is not generally accompanied by morphologically recognizable terminal differentiation. Moreover, with the exception of the GM-CSF-dependent cell line MB-02 ( Perrine et al, 1989 ; Broudy et al, 1993 ), expression from the β-like globin genes is limited principally to the ɛ- and γ- genes with β-globin gene transcription either entirely absent or barely detectable. In contrast, Bb1 and Bb1-3 display a high level of β-like globin gene expression. Indeed, an interesting feature of these lines is the difference in expression of the human β- and γ-globin genes between the original clone, Bb1, and the Bb1-3 derivative which was adapted for growth in a serum-free environment. Although the former expressed almost exclusively human β-globin, the latter expressed equal levels of human γ- and β-globin mRNA. This observation is unlikely to be due to a somatic mutation, since re-adaptation resulted in the same change in globin gene expression, and may therefore reflect an epigenetic change that occurred during the adaptation process. A comparison between the two cell lines in terms of the chromatin structure, methylation status and differences in their trans-acting factor composition may therefore provide valuable information on the developmental regulation of globin gene expression.


We thank Professor Sir David Weatherall for his support, The Wellcome Trust for financial support (Clinical Research Fellowship to A.M.T.), Dr F. Moreau-Gachelin (Paris, France) for the murine spi-1 probe A, Dr J. Jenkins (Norwich) for the mouse p53 cDNA, Dr A. Bernstein (Toronto, Canada) for the mouse fli-1 B1 and B2 probes, Dr S. H. Orkin (Harvard, U.S.A.) for the mouse p45 NF-E2 cDNA, Dr P. Leder (Harvard, U.S.A.) for the mouse c-mpl probe B, Dr S. Gisselbrecht (Paris, France) for the E57 BS probe, Dr C. Perez-Stable (Miami, U.S.A.) for pT7/T3-18βh1, Dr R. Jones (Oxford) for pSPJMαS, Dr M. Baron (Harvard, U.S.A.) for the pSP64Mζ, pSP64Mβ, pSP64Mɛ and pSP64α132 probes, Dr U. Nudel, (Rehovot, Israel) for pAM 19; Nikla Emambokus (Oxford) for pGPIIb1.4RI, Dr Jon Frampton (Oxford) for analysing the DNA content of Bb1-3, and Jackie Sloane Stanley for performing the karyotype analysis.