Cell, Vol. 46, 19-29, July 4, 1966, Copyright 0 1966 by Cell Press Retroviruses as Probes for Mammalian Development: Allocation of Cells to the Somatic and Germ Cell Lineages Philippe Soriano and Rudolf Jaenisch Whitehead Institute for Biomedical Research Department of Biology Massachusetts Institute of Technology 9 Cambridge Center Cambridge, Massachusetts 02142 Preimplantation mouse embryos were infected with a recombinant retrovirus, which serves as a genetic marker for the progeny of an infected blastomere. Quantitative analysis of proviruses carried in mosaic animals indicated that the molarity of individual pro- viral bands was equal In almost all tissues. The cells that give rise to the embryo proper must therefore in- termingle extensively before final tissue allocation to ensure equal contribution of founder cells to all so- matic tissues. The distribution of molarities of in- dividual proviruses suggested that somatic lineages are derived from at most eight founder cells. About half of the proviruses were present in the germ line and the somatic tissues of mosaic animals, while the remaining proviruses were found either in the germ line or in the somatic tissues, but not in both. Our results suggest that at least three cells form the germ line and are set aside prior to somatic tissue allo- cation. Introduction The preimplantation mouse embryo is easily accessible to experimental manipulation, and early lineage relation- ships in mouse development have been established by directly labeling cells (Johnson, 1981; Balakier and Peder- sen, 1982) or by determining the developmental potential of cells in aggregation or injection chimeras (Mintz, 1974; Gardner, 1978; Beddington, 1983; Rossant, 1984). These studies have revealed that two successive developmental decisions are made during preimplantation development that allocate cells to three different lineages (see Figure 1). At the morula stage, cells are allocated to either the trophectoderm or the inner cell mass, and at the late blastocyst stage, inner cell mass cells are allocated to ei- ther the primitive endoderm lineage or the primitive ecto- derm lineage. The time of allocation of cells to a lineage or tissue is defined as the time at which cell mixing be- tween different lineages or tissue primordia becomes minimal (McLaren, 1976). After implantation, the mammalian embryo is not read- ily accessible to direct experimental intervention, and little information is available regarding the subsequent alloca- tion steps of the cells of the primitive ectoderm to the lin- eages that give rise to the somatic tissues and the germ line. The final body plan of the fetus is established at gas- trulation, when the few embryo founder cells have prolifer- ated to form the egg cylinder, which initially consists of some 500-600 cells. These cells divide rapidly during the next 24 hr of primitive streak development, when the cell numbers increase to 12,000-15,000 (Snow, 1977). At the primitive streak stage, different regions of the egg cylinder have acquired different prospective fates, as revealed by orthotopic grafts (Beddington, 1982) as well as in vitro ex- plantation experiments (Snow, 1981). It is completely un- known, however, how the progeny of the few founder cells of the embryo become distributed to different positions in the egg cylinder when final tissue allocation takes place and how many cells are present in the embryo when these developmental decisions are made. The experiments de- scribed in this paper are relevant to this basic question of early postimplantation development. The aggregation of preimplantation mouse embryos carrying different genetic markers has been the predomi- nant method used to label cells for the analysis of cell lin- eages in postimplantation mammalian development. The distribution of cells of the two parental genotypes in the tissues of the resulting chimeric or allophenic mice has been used to obtain information regarding the number of cells set aside to form the embryo and the different so- matic tissues as well as the time of tissue allocation. The use of aggregation chimeras for deducing cell interactions in normal mammalian development has, however, a num- ber of inherent problems, as discussed previously by others (McLaren, 1972; Lewis et al., 1972). It is not clear, for example, how freely the cells of the double-sized tetraparental embryo intermingle after aggregation and whether the reduction in size, which occurs prior to gastru- lation (Buehr and McLaren, 1974; Lewis and Rossant, 1982) distorts normal cell-cell interactions. It furthermore has been observed frequently that one parental genotype is selected over the other during the postnatal life of the chimera (Mintz, 1968; Gearhart and Oster-Granite, 1981). This genotype selection, designated chimeric drift (Warner et al., 1977) also may occur during embryonic development and result in distortions of normal develop- mental processes. Chimeric drift was not observed in chi- meras derived from congenic strains (Behringer et al., 1984), underlining the importance of cell selection in deter- mining the genotypic composition. Cell selection in aggre- gation chimeras is therefore likely to cause problems in deducing lineage relationships for normal mouse devel- opment. The analysis of female X mosaics, while not dis- torting normal development, gives information only on developmental processes subsequent to the time of X-inac- tivation (Nesbitt, 1971). Using this method, McMahon et al. (1983) have suggested that the germ line is allocated after allocation of the somatic lineages. Clearly, for cell lineage studies, a marker for individual progenitor cells that leaves the embryo undisturbed is re- quired. Infection of preimplantation mouse embryos with Moloney murine leukemia virus (M-MuLV) results in effi- cient integration of proviral copies into the host genome. These proviruses can then be used as genetic markers to