Friend acute erythroleukemias are caused in susceptible mice by various strains of Friend leukemia viruses (FLV). The description of the initial isolate took place 50 years ago when the American virologist Charlotte Friend observed that newborn mice inoculated with Ehrlich murine carcinoma cells developed an acute erythroblastosis that could be subsequently transmitted by cell-free extracts prepared from the spleens of diseased mice 20. The early description of the disease indicated that the original isolate (termed anemia-inducing Friend virus or FLV-A) caused an anemia mainly due to hemodilution 21. Since the original report by C. Friend, a number of other isolates capable of inducing a disease spectrum comparable to FLV-A have been identified. Among these, the Rauscher isolate was biologically similar to FLV-A 22. A few years later, derivatives were obtained. In contrast to FLV-A causing an anemia, the derivatives increased the levels of erythrocytes in the peripheral blood (polycythemia) and was termed polycythemia-inducing Friend virus or FLV-P 23. Adult or newborn mice inoculated with either FLV-A or FLV-P develop a disease with no latency. Both virus isolates induce an erythroblastosis that rapidly progresses to acute transformation (Figure 1). The early stage is characterized by a rapid increase in the number of proerythroblastic cells leading to massive hepatosplenomegaly that may be fatal by accidental spleen rupture. These proerythroblasts die in situ or differentiate into abnormal erythrocytes either in the absence of erythropoietin (Epo), in the case of FLV-P, or with a hypersensitivity to Epo, in the case of FLV-A24252627. The target cell in which both FLV-P and FLV-A express their pathogenic effect is an Epo-responsive progenitor cell that was identified as a late BFU-E or a CFU-E 252728. Although Friend virus-infected erythroid progenitors dramatically increase in number, they retain limited proliferation capacity and are not tumorigenic in vivo. Within two to three weeks after virus inoculation, a clonal population of proerythroblastic cells blocked in their differentiation program first emerges in the spleen and then invades the animal 293031. These leukemic cells have extensive proliferative capacity, the ability to be serially transplantable in vivo, and the capacity to eventually establish permanent cell lines in vitro. Cell lines established in culture were called Friend tumor cells. Of note, the erythroid differentiation program can be reinitiated in Friend tumor cells when exposed in vitro to some chemical inducers like dimethyl sulfoxyde, hexamethylene bisacetamide, butyric acid or hemin 32. This contrasts importantly with B-lymphocyte or T-lymphocyte transformation by Bovine Leukemia Virus (BLV) or Human T-cell Leukemia Virus (HTLV)33 which are phenotypically irreversible. Thus, the Friend tumor cells have provided a useful in vitro model for studying the molecular processes of erythroid terminal differentiation.

Both FLV-A and FLV-P isolates contain at least two viral components: a replication-competent Friend murine leukemia virus (F-MuLV) and a replication-defective spleen focus forming virus (SFFV). F-MuLV behaves as a helper supplying the replication-defective functions to SFFV, while SFFV is the pathogenic component responsible for the acute erythroblastosis. SFFV is responsible for the state of either polycythemia (SFFV_P) or anemia (SFFV_A). Further studies have demonstrated that the Rauscher SFFV (Ra-SFFV) was genetically equivalent to the Friend SFFV_A3435. The conclusive proof that SFFV was the pathogenic component responsible for the acute erythroblastosis was that F-MuLV-free viral preparations containing only the SFFV genome could provoke an erythroid proliferative disease when inoculated in mice 36.

The SFFV genome differs from the genome of other acutely oncogenic retroviruses in that it does not contain oncogenic sequences derived from a cellular proto-oncogene. This property is similar to that observed for HTLV-13738. The pathogenicity of SFFV stems from the product of its env gene that encodes a glycoprotein with a molecular mass of 55 kDa (gp55_P and gp55_A) 3940. The env gene contains ecotropic env sequences derived from F-MuLV and dual-tropic sequences probably acquired by genetic recombination with env sequences from an endogenous polytropic virus present as multiple copies in the mouse genome 41. Thus, gp55 presents structural elements homologous to polytropic env glycoproteins in its amino-terminal region and elements homologous to the ecotropic MuLV env gp70 in its carboxy-terminal part. The gp55 contains an extracellular domain, a transmembrane domain and a short cytoplasmic tail 42. Gp55 is not processed in the viral particle but remains in the membranes of the infected cells. A small fraction of gp55 (3–5%) is highly glycosylated in a gp65 form found anchored at the cell surface, while the majority of gp55 remains in the rough endoplasmic reticulum 34. This glycosylation processing is required for the SFFV leukemogenic activity 43.

To decipher the exact role of gp55 in the Friend disease, various models have been elaborated. A retroviral vector transducing only the gene encoding gp55_P was constructed. Upon inoculation in mice, it induced the early stage of the Friend disease similarly to that achieved with the entire SFFV_P genome 44. Another study used transgenesis. The gp55_P-transgenic mice developed a polycythemia associated with a massive proliferation of erythroid precursor cells that were not transplantable 45. Thus, these elegant approaches provided convincing evidence that gp55 is directly responsible for the early erythroblastosis.

An important insight into the gp55 function in erythroid precursor cells was the demonstration that gp55 directly interacts with the receptor for erythropoietin (EpoR) 46. This interaction occurs through the respective transmembrane domains of EpoR and gp55. The binding of gp55 to EpoR, that is highly expressed at the transition of late BFU-E to CFU-E, results in EpoR activation and promotes the Epo-independent proliferation and differentiation of erythroid progenitor cells 42434748. Rare amino acid differences have been identified in the transmembrane domains of gp55_P and gp55_A. Due to these differences, the level of EpoR activation remains incomplete with gp55_A explaining the increase in Epo sensitivity 49 but not autonomy from Epo. In addition to EpoR, gp55 also recruits the short form of the stem-cell-kinase receptor (sf-Stk) as a signaling partner 5051. Stk/RON is a tyrosine kinase receptor of the Met family. sf-Stk retains the transmembrane and the tyrosine kinase domains, but lacks the extracellular ligand binding domain of Stk 515253. Stk/RON is encoded by the Fv2 gene (Friend virus susceptibility gene 2) 52, previously known as a gene determining the susceptibility of some strains of mice (Fv2^s/s) to the Friend disease 5455. Transcription of sf-Stk is naturally initiated from an internal promoter within the Stk gene, and this internal promoter is deleted in mice resistant to Friend virus (Fv2^r/r) 52. Strikingly, the pathology that developed in gp55_P-transgenic mice was not observed under the genetic Fv2^r/r background 45, emphasizing the essential role of sf-Stk in gp55-mediated erythroid proliferation. In SFFV-infected mice, gp55, EpoR and sf-Stk are effectors in signaling pathways which encompass the signal transducers and transcriptional activators STATs 5657, PI3-kinase/AKT 50, the molecular adaptors Grb2/Gab2 58, the Lyn kinase 59, the p38MAP kinase 60, and the ERK1/2 MAP kinases 6162. Constitutive activation of these signaling pathways by the gp55/EpoR and gp55/sf-Stk complexes leads to the dysregulation of proliferation, survival and differentiation of SFFV_P-infected erythroid progenitors in Fv2^s/s mice which result in acute erythroblastosis.

Although SFFV is an acutely transforming retrovirus, it induces malignant transformation of erythroid cells by a mechanism of insertional mutagenesis which is usually employed by non-acute leukemogenic virus. Identification of SFFV proviruses integration sites in the genome of various Friend tumor cells was the first molecular proof for the clonal nature of the leukemic transformation 63. The molecular characterization of the SFFV integration sites revealed that SFFV_P or SFFV_A preferentially integrates at the same genomic locus in 95% of the Friend tumor cells, called spi-1 for SFFV proviral integration site-1 6465. The spi-1 locus is transcriptionally activated by the transcriptional enhancers present in the SFFV LTR 6667, producing the overexpression of a normal spi-1 mRNA translated into a normal Spi-1 protein 6668. During DMSO-induced erythroid differentiation of Friend tumor cells, it was noticed that one of the early events was a marked decline in the level of Spi-1 expression 69. Based on this observation, it was proposed that the arrest in erythroid differentiation was, at least in part, due to Spi-1 overexpression. The high recurrence of Spi-1 insertional mutation in Friend tumor cells suggested that Spi-1 overexpression was cooperative with the constitutive signaling from gp55/EpoR and sf-Stk/EpoR complexes to induce the malignant transformation of the proerythroblast. Indeed, in a heterologous avian model of erythroid self-renewal and differentiation, the effects of Spi-1 on proliferation, survival and differentiation arrest required the co-expression of Spi-1 with an EpoR activated either by gp55 or by a mutation on the residue R129C, which mimics EpoR/gp55 activation 707172.

Besides Spi-1 overexpression, recurrent alterations in the tumor suppressor p53 gene were also detected in the late stage of the Friend disease. Allelic deletions or missense mutations led to loss of p53 tumor suppressive function 737475, and it was shown that Friend leukemia develop more rapidly in transgenic mice expressing a mutant p53 allele, or in p53-null mice, than in normal mice 7677. Although the ectopic expression of a normal p53 protein in Friend tumor cells induces apoptosis and hemoglobin production 78, the loss or the mutation of p53 seems to favor the growth and survival of leukemic proerythroblasts rather than specifically altering erythroid differentiation.

To specify the role of Spi-1 in erythroleukemia, a murine model of spi-1 transgenesis has been constructed 79. The spi-1 transgene, placed under the control of the SFFV_P LTR, was introduced into the germinal lineage of Fv2^s/s mice. Spi-1 transgenic mice develop, within 3 to 4 months after birth, a severe anemia and a massive swelling of the spleen followed by an enlargement of the liver. Bone marrow, spleen, and liver are infiltrated by proerythroblasts arrested in differentiation, mostly at the basophilic stage. These proerythroblastic cells can be established in vitro as permanent cell lines that are exquisitely dependent on Epo for proliferation and survival, and the cells are unable to induce tumors when engrafted in vivo. The Epo dependency was confirmed in vivo by the demonstration that elevation of hematocrit levels (above 60%) by repeated red blood cells transfusions caused regression of the hepatosplenomegaly and disappearance of circulating blasts. Nonetheless, in the animal, erythroblasts expansion occurs in both the presence of Epo in the plasma and the secretion of SCF (stem cell factor) in the marrow or spleen microenvironment. In fact, SCF cooperates with Epo to maintain the in vitro survival and proliferation of the spi-1 transgenic proerythroblasts when Epo is used at limiting concentrations 80. Thus, the spi-1 transgenic model demonstrates that an ectopic expression of Spi-1 results in a block in the maturation program of erythroid precursor cells without removing their growth factor requirement for survival.

The function of Spi-1 in the erythroid differentiation blockage is multifaceted. The extinction of Spi-1 by spi-1-interfering RNAs in spi-1 transgenic proerythroblasts is sufficient to reinstate the erythroid differentiation program 81. This differentiation process is associated with an arrest in proliferation due to both cell death by apoptosis and cell cycle arrest. The mechanisms controlled by Spi-1 that permit survival of the proerythroblast and avoid an exit from the cell cycle are still subjected to speculation. Spi-1 is identical to the transcription factor PU.1, a member of the ETS family. Studies with PU.1 knock-out mice have shown that PU.1 supports hematopoiesis at different stages. Deficiency in PU.1 results in the selective loss of B lymphoid and macrophage development, and a delayed development of neutrophils and T cells 8283848586. In addition, PU.1 plays a role in regulating the commitment of multipotent hematopoietic progenitors 8788. Spi-1/PU.1 functions require the ability of the protein to bind purine-rich DNA sequences in the promoters and enhancers of target genes 6889. Distinct threshold levels of PU.1 also determine function, with high levels driving precursors to a myeloid cell fate, while moderate levels specify B lymphoid development 90. Notably, Spi-1/PU.1 down-regulation is required for normal erythroid development 9192. Thus, Spi-1 overexpression in the proerythroblast probably changes the delicate balance of transcriptional activities required for normal erythropoiesis.

It was postulated that excess Spi-1 may disrupt the function of an erythroid factor and, in this hypothesis, GATA-1 was an obvious candidate 939495. Indeed, GATA-1-deficient erythroid cells fail to mature beyond the proerythroblast stage and die by apoptosis 9697. Several studies have reported a reciprocal inhibition of Spi-1/PU.1 and GATA-1 functions through direct interaction 9899100101. Furthermore, the concept that Spi-1 could inhibit the function of GATA-1 in erythroleukemic cells is supported by the reversal of tumorigenicity and the re-initiation of a differentiation program when GATA-1 expression is ectopically imposed in a Friend tumor cell line 102. Nonetheless, target genes for GATA-1, such as EKLF, NF-E2, β globin, EpoR and GATA-1-itself are expressed in both Friend and spi-1 transgenic cells 79(our unpublished data) indicating that some GATA-1 functions are not abolished by Spi-1 overexpression. Global transcriptome analyses of preleukemic spi-1 transgenic proerythroblasts should provide data allowing the elucidation of the direct targets downstream Spi-1 and the molecular pathways leading to erythroleukemogenesis. As a first indication, transcription of the fli-1 gene is directly regulated by Spi-1 in Friend tumor cells 103. Fli-1 is a transcription factor of the ETS family that behaves as an oncogene when activated by insertional mutagenesis in late erythroid tumors induced by the Friend MuLV alone 104. Thus, Fli-1 dysregulation may contribute, in part, to the erythroid differentiation arrest induced by Spi-1. Another mechanism may involve the phosphatase SHP-1 105 that is transcriptionally up regulated by Spi-1 106. SHP-1 may interfere with the activation of STAT1 and STAT3 and inhibit their DNA binding activities, contributing to the erythroid differentiation arrest57.

During disease progression, fully transformed proerythroblastic cells emerge that are characterized by both autonomous growth in vitro and tumorigenicity in vivo. Thus, like the disease induced by the Friend virus complex, spi-1 transgenic mice develop a malignant process that evolves in two steps (Figure 2) whereby the blastic phase reflects the outgrowth of a malignant erythroblastic cell subpopulation with acquired genetic lesions. According to the data accumulated with studies on the Friend disease, abnormalities occurring in the p53 gene were examined. Inactivating p53 mutations were frequently observed in malignant spi-1 transgenic proerythroblasts, and we reported that disease progression was highly accelerated in a p53-null background 107. Nevertheless, Epo-dependent and non-tumorigenic cells can be isolated during the early erythroblastic phase of diseased p53^-/--spi-1 transgenic mice indicating that p53 germline deletion is not sufficient to confer the malignant phenotype in this context 108. As for the Friend disease, p53 abnormalities appear more as a permissive event supporting the illegitimate survival of proerythroblasts harboring genetic aberrations than a direct transformation event.

Next, signaling alterations were prospected. These studies led to the identification of point mutations in the Kit gene in 86% of tumors isolated late during leukemia progression 80. Kit, the tyrosine kinase receptor for SCF, is expressed on hematopoietic stem cells and committed progenitor cells of the different blood cell lineages 109110. Kit is activated by SCF binding, which induces receptor dimerization followed by the activation of the intrinsic tyrosine kinase and receptor transphosphorylation on specific tyrosine residues 111112. By triggering multiple signaling pathways including ERK1/2 MAP kinases, PI3Kinase, and Src kinases, Kit activation modulates cell survival and proliferation 113114115. Most Kit mutations found in spi-1 transgenic leukemic cells affect amino acids located in the Kit catalytic domain (mainly codon 814 and occasionally codon 818). Of note, similar mutations were found in human mastocytosis 116117 and acute myeloid leukemia 118119120. These gain-of-function mutations confer ligand-independent tyrosine kinase activity to Kit. The mutated forms of Kit constitutively activate MAP kinases and PI3Kinase/AKT pathways in the spi-1 transgenic leukemic cells 80121. When the expression of Kit mutants is enforced in spi-1 transgenic preleukemic proerythroblasts, cells become growth-factor independent and tumorigenic in vivo. Thus, a constitutive signaling from Kit mutants combined to Spi-1 overexpression leads to malignant transformation of proerythroblasts (Figure 3).

AML in human are characterized by an uncontrolled expansion of immature blasts, which fail to differentiate normally. These leukemias are classified into sub-types according to the predominance of the altered myeloid lineage. It is currently accepted that the disease progresses from a chronic phase to an acute blastic crisis. Molecularly, leukemias are associated with multiple genetic alterations, including chromosomal translocations and mutations. Overviews on the genetic alterations found in AML are detailed in recent reviews 122123124125126127. One important observation is that the genetic alterations can be divided into two classes. One class targets transcription factors that play a regulatory role in hematopoietic development. These transcription factors are frequently modified through chromosomal translocations and the resulting chimeric proteins inhibit differentiation in a particular hematopoietic cell lineage. For example, the promyelocytic leukemia-retinoic acid receptor α (PML/RARα) expressed from the t(15;17) translocation is associated with 99% of acute promyelocytic leukemia 128. PML/RARα behaves as a transcriptional repressor for the hormonal receptor RARα 129. In the second class, mutations affect signaling effectors like tyrosine kinases 125. These mutations are gain-of-function and provide proliferative signals, as exemplified by point mutations inducing the constitutive activation of the tyrosine kinase receptor FLT3 in ~30% of AML 122130131. Strikingly, murine models designed to express one genetically modified protein, using the transplantation of retrovirally-transduced bone marrow cell or transgenesis, showed that one type of mutation does not induce an acute leukemia on its own. Generally, they provoke a myeloproliferative disorder. Progression to leukemia requires either a long latency, making plausible the occurrence of a second mutation, or an additional treatment with mutagenic compounds. The concept emerging from this complexity is that an acute leukemia arises from the cooperation between two events: one mutation interfering with differentiation and the other mutation conferring a proliferation advantage 132.

Both the Friend and spi-1 transgenic models demonstrate that leukemia development depends on the cooperation between one mutation that impairs differentiation and a second mutation that promotes autonomous cell growth (Figure 1). In that way, these murine models provide proof-of-principle for the "two-hits" model of leukemogenesis 132. They highlight that complementation between the two classes of mutations is more important for leukemogenesis than the timing order of their occurrence.