The Prolactin Receptor and Severely Truncated Erythropoietin Receptors Support Differentiation of Erythroid Progenitors*

Activation of the erythropoietin receptor is essential for the survival, proliferation, and differentiation of erythroid progenitors. To understand the role of erythropoietin receptor (EpoR) activation in erythroid differentiation, we infected primary erythroid progenitors with high-titer retrovirus encoding the non-hematopoietic prolactin receptor. The infected progenitors responded to prolactin in the absence of Epo by generating fully differentiated erythroid colonies. Therefore, differentiation of erythroid progenitors does not require an intracellular signal generated uniquely by the EpoR; the EpoR does not have an instructive role in erythroid differentiation. We also infected primary erythroid progenitors with retrovirus encoding chimeric receptors containing the extracellular domain of PrlR and the intracellular domain of either the wild-type or truncated EpoRs. A chimeric receptor containing only the membrane-proximal 136 amino acids of the EpoR cytoplasmic domain efficiently supported prolactin-dependent differentiation of erythroid progenitors. Substitution of the single cytoplasmic domain tyrosine in this receptor with phenylalanine (Y343F) eliminated its ability to support differentiation. The minimal EpoR cytoplasmic domain required for erythroid differentiation is therefore the same as that previously reported to be sufficient to support cell proliferation (D’Andrea, A. D., Yoshimura, A., Youssoufian, H., Zon, L. I., Koo, J. W., and Lodish, H. F. (1991) Mol. Cell. Biol. 11, 1980–1987; Miura, O., D’Andrea, A. D., Kabat, D., and Ihle, J. N. (1991) Mol. Cell. Biol. 11, 4895–4902; He, T.-C., Jiang, N., Zhuang, H., Quelle, D. E., and Wojchowski, D. M. (1994) J. Biol. Chem. 269, 18291–18294).

The EpoR 1 belongs to a large family of cytokine receptors, many of which are required for the proliferation and differentiation of hematopoietic as well as other cell types (4 -7). Throughout life, eight different hematopoietic lineages arise from pluripotent stem cells in the bone marrow (8,9). The exact role of growth factors in this process is not clear and has been described broadly by two alternative hypotheses. The stochastic hypothesis suggests that commitment of a progenitor to a particular lineage is a stochastic event, subsequent to which cell differentiation proceeds along a predetermined program; growth factors are merely required to ensure the survival and proliferation of committed progenitors (10 -14). The contrasting inductive, or instructive, hypothesis attributes to growth factors a direct role in cell differentiation, predicting that growth factor receptors transduce signals that uniquely specify the differentiation outcome of a progenitor (15)(16)(17). A number of hybrid hypotheses have also been proposed, where, for example, committed progenitors arise stochastically, but their subsequent differentiation and acquisition of the mature phenotype are uniquely induced by lineage-specific growth factors (18).
Although Epo is essential for the production of red blood cells, it is not thought to play a role in the generation of committed erythroid progenitors from pluripotent progenitors: expression of recombinant EpoR does not bias the lineage commitment of pluripotent hematopoietic progenitors (19,20), and normal numbers of committed erythroid BFU-e and CFU-e progenitors are found in the fetal livers of EpoR Ϫ/Ϫ mutant mice (21). However, there is a unique requirement for EpoR activation during the subsequent proliferation and terminal differentiation of committed erythroid progenitors: the EpoR Ϫ/Ϫ CFU-e and BFU-e progenitors fail to give rise to mature erythrocytes unless EpoR is expressed by retroviral infection (21); and in vitro, other growth factors only partially substitute for Epo (22)(23)(24). It is not known whether EpoR activation is essential at this stage of erythroid differentiation because of an EpoR-unique instructive signal or whether it is simply required for functions that are not unique to EpoR, such as its proliferative and anti-apoptotic effects.
Some evidence for the capability of EpoR to promote the erythroid phenotype comes from the ability of Epo to induce surface expression of glycophorin (25) and transcription of the ␤-globin gene (26,27) in pre-B Ba/F3 cells expressing a transfected EpoR. However, the uncertain lineage commitment of many cell lines and their incomplete differentiation response makes them less suitable for the study of signaling in differentiation. EpoR-mediated signaling for proliferation can be studied in a number of interleukin-3-dependent cell lines, where heterologous expression of EpoR allows Epo to support cell proliferation (1,2). Only the membrane proximal ϳ120 amino acids is essential for this function (1)(2)(3). Similarly truncated mutants of other cytokine receptors are also able to support mitogenic signaling in such cells (17, 28 -30). Since the greatest homology between cytokine receptors is found in the Box 1 and Box 2 domains of their membrane-proximal regions (30) (see Fig. 1), it might be expected that this region would generate signals for functions common to all these receptors such as cell survival and proliferation and that the divergent membranedistal regions would endow the specificity of signaling presumed unique to each receptor.
We therefore examined whether the distal half of the EpoR cytoplasmic domain is essential for differentiation of primary fetal liver erythroid progenitors. We also examined whether the entirety of the cytosolic domain of the EpoR can be replaced with the corresponding segment of a different receptor; we chose the prolactin receptor, which plays no role in hematopoiesis, but belongs to the same subfamily of cytokine receptors as EpoR, and shares many of its signaling molecules.

EXPERIMENTAL PROCEDURES
Chimeric Receptors-CHI338 was generated by ligating a doublestranded oligomer containing a stop-codon in frame at position 339 of the EpoR between the BSP120I and HindIII sites of the murine EpoR cDNA (31). CHI374 was made by inserting a stop-codon in frame in the HindIII site of CHI. CHI374/Y343F was constructed by ligating a polymerase chain reaction product containing the point mutation into the BSP120I and HindIII sites of CHI374. CHI442 was made by subcloning the BSP120I-EcoRI fragment of pSFFV.tEpoR (32) into the BSP120I site of CHI. All receptors were expressed in the retroviral expression vector MSCV (33).
Transducing Retroviruses-The 293 cell line expressing the MuLv gag-pol proteins (34) was co-transfected with p env.min encoding the mouse ecotropic MuLv envelope glycoprotein, a gift of Dr. David A. Sanders (Purdue University), and with p SV2neo. Fifty G418-resistant clones were selected; of these, clone VE23 generated the highest titer of transducing retroviruses following transient transfection with a reporter MSCV DNA containing the lacZ gene. To generate transducing retroviruses, VE23 cells were transiently transfected using the calciumphosphate method with MSCV retroviral constructs each encoding the desired receptor. Culture supernatants were collected at 48 h and either immediately frozen or used for infection.
Fetal Liver Cell Infection and Culture-Fetal livers from BALB/c mouse embryos were harvested at days E13 to 15 and dissociated by mechanical pipetting. The cells were subjected to a brief treatment with distilled water to lyse non-nucleated cells, strained through a 70-micron cell filter, and washed in Iscove's modified Dulbecco's medium (IMDM; purchased from Life Technologies Inc.) and 20% fetal calf serum. The cells were resuspended either in virus-containing culture supernatants containing 4 g/ml polybrene or in control culture medium with 4 g/ml polybrene. After rocking for 3-4 h at 37°C, the cells were recovered by centrifugation. For CFU-e cultures, the cells were washed in IMDM and resuspended in serum-free methylcellulose medium (1% methylcellulose in IMDM, 1% bovine serum albumin, 100 ng/ml IGF-I, 7.7 g/ml cholesterol, 5.6 g/ml oleic acid, 8 g/ml L-␣-phosphatidylcholine, 0.3 mg/ml transferrin, 4 l/100 ml monothioglycerol) and ovine-prolactin (the National Hormone and Pituitary Program of the NIDDK, National Institutes of Health, Bethesda, MD) as indicated. For BFU-e cultures, they were resuspended in 30% fetal calf methylcellulose medium (Stem Cell Technologies) with added spleen conditioned medium, 50 ng/ml rat stem cell factor, and 500 ng/ml ovine prolactin. Control (uninfected) cultures in each experiment had Epo added at 2 units/ml. Hemoglobinized CFU-e colonies were scored on day 3 after staining with diaminobenzidine. Hemoglobinized BFU-e colonies were scored on day 7.
FACS Scanning-A small aliquot of cells from each infection was cultured in 20% fetal calf serum with 2 units/ml Epo and used at 36 h. The cells were washed three times in phosphate-buffered saline containing 0.5% BSA, incubated at 4°C with 50 g/ml monoclonal antibody M110 (gift of Dr. J. Dijane) in the presence of 200 g/ml each of rabbit and goat IgG, followed by a phycoerythrin-conjugated goat anti-mouse F(abЈ) 2 . Cells were scanned on a Becton-Dickinson cell scanner.
Cytospins-Five day 7 BFU-e bursts were individually aspirated from each methylcellulose culture. Following cytospin, cells were fixed in methanol, stained in diaminobenzidine, and counterstained with Giemsa.  Fig. 1). We then examined the ability of the infected fetal liver cells to generate erythroid colonies when cultured in semi-solid methylcellulose medium in the presence of prolactin (Fig. 2). In parallel, we measured the level of expression of the chimeric receptors by culturing a small aliquot of the infected fetal liver cells in liquid medium in the presence of Epo. After 36 h we incubated the cells with the monoclonal antibody M110, specific for the prolactin receptor (36), and quantified the fraction of M110-positive cells by FACS. The window we employed (M1, Fig. 2A) provides a lower estimate of the fraction of fetal liver cells infected by each of the chimeric receptors. Infection by CHI varied from 10 to 25% (Figs. 2, A, C, and D).

RESULTS AND DISCUSSION
In each of five experiments, the ratio of prolactin-dependent erythroid BFU-e and CFU-e colonies formed by CHI-infected cultures to those formed by parallel uninfected and erythropoietin-treated cultures was the same or higher than the estimated rate of infection by CHI (Fig. 2). Specifically, Epo supported the generation of 700 -1000 CFU-e colonies (per 2 ϫ 10 5 fetal liver cells; C); 10% of the fetal liver cells expressed CHI, and this population generated 100 CFU-e colonies promoted by prolactin. Thus the chimeric receptor CHI, containing the full-length cytoplasmic domain of EpoR, functions as efficiently as the endogenous EpoR in supporting erythroid colony development.

CHI374 Contains a Minimal EpoR Domain Required to Efficiently Support Erythroid Differentiation; Tyr-343 Is Essential for This Function-CHI442 and CHI374
were each able to support day 3 prolactin-dependent erythroid colonies in serumpoor medium (CFU-e-derived colonies) and day 7 erythroid bursts (BFU-e-derived bursts) to the same extent as CHI (Fig.  2, B and C). CHI338 was not able to support erythroid colony formation, although it was well expressed by the infected fetal liver cells (Fig. 2, B and D). No significant quantitative differences were seen between CHI, CHI442, or CHI374 in their response to prolactin (Fig. 2, B and C). The reduction in numbers of colonies and bursts at high concentrations of prolactin (Fig. 2, C and D) is expected, assuming the PrlR, a member of the same family of cytokine receptors as the Epo and growth hormone receptors, follows the same sequential dimerization mechanism (37). Importantly, the colony and burst size, mor- phology, and degree of hemoglobinization promoted by CHI, CHI442, and CHI374 were similar to those promoted by Epo acting through the endogenous EpoR (Fig. 3). Therefore, CHI374 contains a minimal EpoR domain required for erythroid differentiation. Tyrosine 343 within this domain is essential for erythroid differentiation, since its replacement in CHI374 with phenylalanine completely abolished its ability to support CFU-e differentiation (CHI374/Y343F, Fig. 2D). This Burst numbers represent mean Ϯ S.E. of triplicate measurements, scored on day 7 of culture. In parallel, and in the same experiment, uninfected control methylcellulose cultures incubated with 2 units/ml Epo generated 110 bursts per 2 ϫ 10 5 cells. C and D, prolactin dependence of CFU-e-derived colonies in serum-poor methylcellulose medium. Two representative experiments are shown out of four; each receptor was tested in at least two experiments. Shown for each receptor is the fraction of infected fetal liver cells, as determined by culturing cells for 36 h postinfection in 2 units/ml Epo and staining with fluorescent anti-PrlR monoclonal antibody M110. In parallel, and in the same experiment, uninfected control cultures incubated in serum-free methylcellulose medium with 2 units/ml Epo generated between 700 and 1000 CFU-e colonies/2 ϫ 10 5 cells. E, PrlR is as efficient as CHI in supporting BFU-e burst formation. Each determination is the mean Ϯ S.E. of three independent experiments. The mean fraction of M110positive cells was 8% for the PrlR and 9% for CHI. In parallel, uninfected control cultures incubated in serum-free medium with 2 units/ml Epo generated an average of 90 bursts per 2 ϫ 10 5 cells. same tyrosine is essential for mitogenesis and is one of the four in the EpoR able to mediate STAT5 activation (38 -41). The minimal EpoR cytoplasmic domain required to support erythroid differentiation is therefore the same as that previously found to be sufficient to support cell proliferation (1)(2)(3).
Wild-type PrlR Can Fully Support Differentiation of Erythroid Progenitors-The finding that CHI374 can fully support erythroid differentiation suggests that there is no essential differentiation signal emanating from the distal part of EpoR; if there is such a differentiation signal, it must arise from the membrane-proximal part of the receptor. To examine this point further, we infected progenitors with retrovirus encoding the wild-type PrlR. Like the EpoR, activation of the PrlR leads to Jak2 and STAT5 phosphorylation (35,42,43). In hematopoietic cell lines such as Ba/F3, transfected PrlR supports Prl-mediated cell proliferation (35,42). Figs. 2, C and D, and 3 show that the PrlR is as efficient as CHI in supporting formation of both CFU-e-and BFU-e-derived erythroid colonies. CHI and PrlR were expressed at similar levels by the fetal liver cells, and there was no significant difference between the prolactin responsiveness of the CHI-infected and PrlR-infected cells (Fig. 2, C and D). PrlR-supported colonies were of the same size, morphology, and extent of hemoglobinization as those that arose from prolactin-dependent CHI-infected progenitors or control uninfected Epo-dependent progenitors. Normal erythrocytes were present in cytospin preparations of PrlR-supported colonies (Fig. 3). Thus, the cytosolic domains of the Epo and Prl receptors were equally efficient in supporting erythroid differentiation when the extracellular domain was activated by dimerization by prolactin. There is therefore no requirement for an EpoR-unique signal during the differentiation of erythroid progenitors.
The unique phenotypic outcome of erythroid differentiation cannot therefore be due to an essential, EpoR-generated instructive signal. Erythroid differentiation from BFU-e and CFU-e progenitors apparently proceeds along a predetermined program, supported by EpoR-activated "generic" intracellular signals also common to other cytokine receptors. The exact role of these EpoR-activated signal transduction proteins is still to be determined. Our finding that the minimal domain of EpoR able to support differentiation is the same as that required to support cell proliferation raises the possibility that the role of the EpoR-activated signals is only to support the survival and proliferation of the differentiating progenitors, as proposed by the stochastic hypothesis. It is possible, however, by analogy with studies on the granulocyte-colony-stimulating factor or thrombopoietin receptors (44,45), that, to support erythroid gene expression, additional signals are needed beyond those required for proliferation. If this is the case, these signals must arise from the membrane-proximal domain of the EpoR and cannot be unique to the EpoR. The inability of tyrosine kinase receptors to fully replace EpoR function (22,23,46) suggests that the signals required for erythropoiesis, although not unique to the EpoR, are not shared by all receptors.
Relatively little is known about how unique patterns of gene expression are generated by different cytokine ligands, particularly since many of the known signaling proteins are common to more than one receptor (47,48). All of the signaling proteins known to be activated by the EpoR, including JAK2, STAT5, SH-PTP1, Grb2, and Ras, are also activated by other cytokine receptors (6,7). Here we show that the unique outcome of EpoR signaling is not due to any unique signals it may generate, but instead is encoded in the cellular environment of erythroid progenitors. The specificity of EpoR function in erythroid differentiation is therefore a result of its unique expression by erythroid progenitors.