Generation of pigmented stripes in albino mice by retroviral marking of neural crest melanoblasts

The classical tissue grafting studies of Rawles (1940, 1947) established that the pigment cells of mammalian skin are derived from the neural crest, a pluripotent cell population that migrates extensively through the midgestation embryo. Neural crest cells arise from the lateral edges of the neural folds during closure of the neural tube, and subsequently disperse from the neural tube in two loosely organized streams of cells (see Le Douarin, 1982 and Weston, 1983, for reviews). One stream penetrates ventrally into the body, giving rise to a variety of cell types, including virtually all the neurons of the peripheral nervous system. The other spreads laterally away from the neural tube and travels around the outer periphery of the embryo towards the ventrum. This is the route taken by melanoblasts, the precursors of pigment-producing melanocytes, which migrate through the mesoderm underlying the skin ectoderm (Mayer, 1973), eventually invading the ectoderm to colonize the skin and developing hair follicles. In the mouse, this migration is initiated on embryonic day (E) 8.5 and completed on E12 with the entry of melanoblasts into the ectoderm (Rawles, 1947).

The lateral migration of melanoblasts away from the dorsal mid-line can be readily discerned in the coat color patterns of murine aggregation chimeras or allophenic mice. These animals, generated by fusion of embryos of two genotypes differing in coat color, are characterized by broad transverse stripes of pigment, often with a sharp line of discontinuity at the dorsal mid-line (Mintz, 1967; McLaren and Bowman, 1969). Mintz (1967) interpreted each ‘standard’, or unit-width, stripe as representing the clonal progeny of a single progenitor melanoblast and concluded from the width of the transverse bands, and the independence of left and right sides, that pigmentation of the coat is derived from 17 pairs of founder melanoblasts arrayed as 17 founder clones per side. Others (Wolpert and Gingell, 1970; West, 1975a) have challenged these conclusions based on the probability of 17 randomly chosen clones giving rise to 17 stripes. McLaren (1969) has also noted that because aggregation chimeras are generated by mixing embryos of two different genotypes, a stripe could, in fact, represent the progeny of several clones if genetically similar cells within the chimera showed a tendency to aggregate during development. A more direct way of deriving a clonal history of melanoblasts, and determining the number of founder cells giving rise to coat pigmentation, is to generate single genotype mosaic animals rather than aggregation chimeras. Indeed, support for Mintz’s model has been provided by the pigmentation patterns of mice heterozygous for X-linked color genes (Cattanach et al. 1972) and by mice carrying mutant color genes with high somatic reversion rates (Mintz, 1970).

One experimental approach for generating single genotype mosaic animals involves genetically labeling single cells of the embryo in situ, in order to allow assessment of the clonal expansion of founder cells in the animal. In previous studies, we have demonstrated the efficacy of in utero microinjection procedures for introducing replication competent retroviruses (Jaenisch, 1980), as well as replication defective retroviral vectors (Compere et al. 1989a,b) into neurulating mouse embryos. In the latter studies, which involved transduction of activated oncogenes, a large fraction of animals developed tumors bearing a single proviral copy of the vector, demonstrating the feasibility of marking single embryonic cells by this approach.

In this study, we have utilized the in utero microinjection technique with the aim of generating mosaic mice with patches of coat pigmentation derived from retrovirally marked founder melanoblasts. A retroviral vector, MLV-HuTyr, bearing a human tyrosinase cDNA under the transcriptional control of the MLV long terminal repeat (LTR), was microinjected into mid-gestation albino embryos. Albino mice are white due to deficiency of tyrosinase, the key enzyme in melanin synthesis, but otherwise contain a functional melanoblast population (Silvers, 1956). Introduction of the wild-type tyrosinase gene into mutant melanoblasts should rescue the albino defect, giving rise to patches of pigmentation corresponding to the area colonized by the mitotic progeny of a marked clone. Thus, this approach should directly visualize the clonal expansion of a single progenitor melanoblast. Our results demonstrate that in utero transduction of neural crest cells is feasible, and that the pigmented bands generated by this approach are very similar in width to the standard stripes observed in aggregation chimeras. In addition, pigmentation was also observed in the eyes of many of the mosaic animals. Unlike coat pigmentation, pigment in the eyes is not derived solely from neural crest melanoblasts. The cells of the pigmented layer of the retina and the inner layer of the iris differentiate locally from the neuroepithelium of the optic cup. Mosaic mice generated by in utero infection with MLV-HuTyr showed pigmentation of both neural crest and CNS-derived tissues in the eye.

BALB/c mice were obtained from Jackson labs and FVB/N mice from Taconic; both strains were bred in our mouse colony.

Female mice were mated with males in the evening, and successful matings were identified the following morning by vaginal plug formation. The day of plug formation was taken as day 0 (E0) of gestation.

NIH 3T3 cells and the retrovirus packaging line GP+E-86 (Markowitz et al. 1988) were grown in Dulbecco modified Eagle medium (DMEM) containing 10% calf serum (CS).

Virus infections were carried out by co-cultivation of approximately 5×10⁵ 3T3 cells, in a 60 mm dish, for 1h with 1 ml of virus-containing medium supplemented with 8 μg ml⁻¹ polybrene. Infected cells were passaged the following day, harvested 3 days after infection, and frozen at − 20°C as dry pellets for subsequent use in tyrosinase assays and for Southern blot analysis.

The MLV-HuTyr vector was constructed by replacing Moloney murine leukemia virus (MoMLV) sequences between the PstI site at nucleotide 563, and the Clal site at 7674 (see Van Beveren et al. 1985 for sequence numbering) with the 1.8 kb EcoRI-A’del fragment of the human tyrosinase cDNA. The cDNA subclone was derived from the expression vector pcTYR (Bouchard et al. 1989), and contained the tyrosinase coding sequences in their entirety but lacked the endogenous polyadenylation site. The vector also carried the B2 point mutation in the tRNA-binding site (Barklis et al. 1986).

For virus production, the vector was cotransfected with pSV2-neo (Southern and Berg, 1982) into GP+E-86 cells, and transfected cells selected by growth in G418 (200 μg ml⁻¹ of free base). Clonal lines of transfected cells were isolated, then frozen, while culture medium was screened for vector production by an RNA dot blot assay (Huszar et al. 1989) using the tyrosinase cDNA as probe. As described in Results, the titer of virus-producing clones was greatly decreased after thawing, thus a pooled transfected population of G418 resistant GP+E-86 cells, which had been grown continuously since transfection, was used as the source of virus. The pooled cells were grown to confluency and incubated overnight with fresh culture medium which was harvested, centrifuged at 10000 g for 20 min at 4 °C, and the supernatant frozen in aliquots at −70 °C.

MLV-HuTyr virus was concentrated essentially as described by Price et al. (1987). Virus-containing medium was centrifuged at 10000g for 16-20 h at 4°C, the supernatant discarded, and the pellet gently solubilized in 1/100 volume DMEM/10% calf serum, and stored in aliquots at −70°C.

Genomic DNA was extracted by vortexing cells into 10 mw Tris-HCl, pH7.4, 10mM EDTA, 10mM NaCl and 100μgml⁻¹ proteinase K. SDS was added to 1 % and samples were incubated at 37 °C for 3h followed by extraction with phenol and precipitation with ethanol. 6 pg of DNA was digested with Sad, electrophoresed through an 0.9% agarose gel and transferred to a nylon membrane (Zetabind, Cuno). Membranes were hybridized to the radiolabelled (Feinberg and Vogelstein, 1983) 2 kb EcoRI human tyrosinase cDNA fragment of pcTYR (Bouchard et al. 1989).

Cells were disrupted by freeze-thawing in PBS/1% NP40, and the crude extract obtained following pelleting of cellular debris was used for enzyme assays. Tyrosinase activity was assayed essentially as described by Pomerantz (1969). Briefly, cell extracts containing 20, 50, 100 and 200 μg of protein were incubated with 5μCiml⁻¹ [³H]tyrosine, 50μM L-tyrosine, and 50 UM L-dihydroxyphenylalanine (L-DOPA), in 200 μl of 50mM phosphate buffer pH 7.0 for 45 min at 37°C. The assay was linear over this time period at the protein concentrations used. Protein concentration was determined using the BioRad protein assay kit (Bio-Rad Laboratories).

Microinjection was performed as previously described (Hus-zar et al. 1991). Briefly, laparotomy was performed on anesthetized pregnant females on E8.75 or 9.75. The uterus was held with forceps while a glass micropipette containing concentrated MLV-HuTyr virus, which had been supplemented with 100μgml⁻¹ polybrene, was inserted into the ventral third of the decidual swelling (at this stage of development the embryo cannot be visualized within the uterus). Each embryo was injected with a total of 0. – 0.5 /d of virus.

Tissues were fixed in 10 % buffered formalin, then dehydrated in graded ethanols and xylenes, and embedded in Paraplast plus ®. Sections 4;xm thick were prepared and stained with Harris’ hematoxylin and eosin.

To generate infectious virus, the MLV-HuTyr vector (Fig. 1) was introduced into GP+E-86 retrovirus packaging cells (Markowitz et al. 1988). Transfected Gp+E-86 cells were pooled, grown as a mass population, and expanded for harvesting virus. Initially, individual clones of transfected cells were also isolated, and then frozen while culture supernatant was screened for MLV-HuTyr production. Upon thawing of virusproducing clones, however, the cells underwent a crisis characterized by a decreased rate of cell growth, the appearance of pigmented cytoplasmic inclusions, and a marked decrease in viral titer (not shown). This was likely a consequence of tyrosinase expression in the GP+E-86 fibroblast cell line, since the cytotoxicity of the products of tyrosine metabolism has been well documented (Pawelek and Lerner, 1978), and presumably, fibroblasts lack the mechanism by which melanocytes protect themselves from physiological levels of the toxic intermediates. Our experience is that these cytotoxic effects are most noticeably manifested following a cycle of freezing and thawing of the producer cells (not shown).

Southern blot analysis of Sad digested genomic DNA from the pooled virus producer cells demonstrated the presence of approximately two proviral vector copies per GP+E-86 cell (Fig. 2A, lane 1), as estimated by comparison of band intensity with copy number standards. The cells expressed high levels of tyrosinase activity (Fig. 2B, lane 1), comparable to that of control human melanoma SK-MEL-1 cells. Production of MLV-HuTyr virus was quantified by incubation of 3T3 cells with 1 ml of culture medium from the pooled producer cells. We observed the integration of approximately 0.2 proviral copies per 3T3 cell (i.e. 1 in 5 cells infected; Fig. 2A, lane 2). In order to increase viral titer for embryo injections, the viral stock was concentrated by centrifugation. As shown in Fig. 2A (lane 3), this resulted in a 10-fold increase in viral titer: incubation of 3T3 cells with a four-fold dilution of concentrated virus gave rise to 0.5 proviral copies per cell and a corresponding increase in enzyme activity over that observed in cells infected with unconcentrated virus (Fig. 2B, lanes 2 and 3). Uninfected 3T3 cells contained no detectable levels of tyrosinase activity (Fig. 2B). Based on the number of 3T3 cells used for infection (∼5×10⁵) the titer of the concentrated virus can be estimated at ∼10⁶ infectious particles per ml.

The concentrated preparation of MLV-HuTyr was used for in utero injection of BALB/c and FVB/N embryos. Since the midgestation embryo cannot be visualized within the uterus, the injection pipette was inserted, and virus deposited, within that portion of the decidua known to house the embryo.

As shown in Table 1, over 60% of the injected animals survived to adulthood. Of those embryos infected on E8.75, 22% (41/188) showed pigmentation of the coat and/or eyes, whereas the success rate was approximately 10-fold lower (2.5 %) on E9.75. Mosaic mice exhibited faint stripes and patches of hair pigmentation after the first week of life. Upon close inspection, many animals also showed evidence of pigmentation in the eyes. Representative pigmented mice are shown in Fig. 3. Hair pigmentation appeared as a light brown color in both BALB/c and FVB/N mice, since these strains cany the wt agouti (A) allele and are also homozygous for the brown (b) allele. Both alleles cause black hair to be of lighter appearance: the agouti allele induces alternating deposition of black eumelanin and yellow phaeomelanin into the hair, whereas the brown mutation produces brownish eumelanin (Silvers, 1979). In addition, insufficient in vivo expression of the tyrosinase cDNA may be a further contributing factor in the faintness of the pigmentation. Interestingly, the intensity of pigment increased gradually over the time of observation (3–4 months) in some of the mosaics. This may indicate a competitive advantage of melanin-producing over albino melanocytes, resulting in a gradual selection for the former in those hair follicles containing both types of melanocytes.

The majority of mosaics (29 of a total of 38 mice with coat pigmentation) showed only one discrete pigmented area, likely attributable to infection of a single melanoblast clone; in the remainder, usually 2 or 3 distinct, non-contiguous, pigmented areas were observed (Table 2). In most of the mosaics, pigmentation was restricted to the head (Table 2), manifested as either a band extending from the dorsal mid-line on the forehead (Fig, 3A), or as a larger patch on the face or head which also delineated the dorsal mid-line (Fig. 3B). In some of these mosaics, the extent of cranial pigmentation was much more limited, consisting of only 2 or 3 pigmented hairs dose to the dorsal midline, or a faint patch made up of several pigmented hairs extensively interspersed with white hairs (not shown).

Pigmentation of the trunk was more uniform than that observed in the head. Pigmented areas always appeared as a stripe, of consistent width, extending ventrally from the dorsal mid-line (Fig. 3C-F). Often these stripes appeared to terminate along the flank of the animal (Fig. 3D,E); however, closer inspection revealed that this was due to dispersion of the pigmented hairs into two or three narrower stripes as well as to a progressive decrease in the number of pigmented hairs (not shown). The progressive diminution of pigment towards the ventrum is also characteristic of the stripes in aggregation chimeras, reflecting the dorsoventral direction of melanoblast migration. For the most part, the pigmented stripes were cleanly delineated at the dorsal mid-line; however, in 2 of the 10 mosaics with trunk pigmentation, pigmented hairs were observed to extend bilaterally from the mid-line around both sides of the mouse (e.g. Fig. 3F). Both of the mosaics derived from injection on E9.75 showed trunk pigmentation which was very faint (i.e. consisting of relatively few pigmented hairs), appearing as 3 short stripes extending only a few millimeters from the dorsal mid-line (not shown).

Microscopically, brown granular pigment was observed throughout the mature hair shafts of pigmented hairs; in cross-section the brown granules could also be seen within hair follicles (Fig. 4B). In those animals with skin pigmentation, brown pigment was seen in stellate cells, presumably melanocytes, present within the dermis of pigmented areas of the eyelids or ears (Fig. 4A).

External inspection frequently revealed pigmentation in the eyes of mosaic mice (Table 1). When killed, the eyes of mice with coat pigmentation and with externally visible pigment in one or both eyes were removed and examined under a dissecting microscope. The intensity of pigmentation was highly variable, ranging from the presence of only a few pigmented cells to extensively pigmented eyes which appeared almost black (Fig. 3G). Two types of pigment cells in the eyes of the mosaics could be distinguished with a dissecting microscope: large stellate cells, typical of neural crest-derived melanocytes, observed in the choroid and the outer layer, or stroma, of the iris (Fig. 3H); and small epithelial cells, not of neural crest origin, typical of the pigment layer of the retina (Fig. 3I). These latter cells were often seen as columns of pigmented cells radiating from the optic nerve, as has also been seen in chimeric mice (West, 1975b).

Light microscopic analysis of representative samples of these pigmented eyes confirmed the assignment of pigment cells to the outer layer of the iris and choroid (neural crest-derived) or to the inner layer of the iris and pigment epithelium of the retina (CNS-derived). Representative examples are shown in Fig. 4C and D. In eyes with retinal pigmentation, brown retinal pigment epithelial cells were either present individually or in small clusters of 3 –10 cells. Multiple clusters of the pigment cells were observed in extensively pigmented eyes. Similarly, regions containing brown pigmented cells were observed in the choroid, in the ciliary epithelium of the ciliary body and the epithelial pigment layer on the posterior surface of the iris (not shown). In heavily pigmented eyes, these cells formed a contiguous layer.

When eye pigmentation in choroid and stroma of the iris was compared with coat pigmentation on the head, a strong positive correlation was observed: 22/28 mice with pigmentation in the head also had pigmentation in choroid and iris stroma and, conversely, 27/30 eyes with pigment in choroid and outer iris were in mice with head pigmentation. In addition, the skin in the eyelids was pigmented in several mice with pigment in choroid and/or outer iris. In contrast, about half of the eyes with retinal pigment (9/19) were seen in mice with no detectable head pigmentation (Table 3). These results are consistent with the developmental origin of the pigment cells: the melanocytes of the skin, the choroid and the iris stroma are all derived from the neural crest, whereas the pigment cells of the retina are of a different lineage, originating from the neuroepithelium. Furthermore, the results suggest that melanocytes in the head area and in the choroid/outer iris are clonally related.

Because pigmented cells of neural crest and non-neural crest origin were observed in eyes of these mice, the mice were further examined for evidence of pigmentation in their internal organs. In particular, it was of interest to determine whether other neural crest-derived cells were pigmented as well as whether other cells capable of producing melanin, but not of neural crest origin (e.g. cells of the substantia nigra) were pigmented. Macroscopic examination of the heart, lungs, thymus, liver, spleen, pancreas, kidneys, gastrointestinal and reproductive tracts revealed no evidence of pigmentation. Nine adrenal glands had areas suggestive of pigmentation macroscopically. Upon microscopic examination, a tan-brown to golden-brown finely dispersed pigment was observed in the adrenal medulla of three adrenal glands. It was not clear whether this pigment was lipofuscin or melanin. Brains of five mosaics with extensive pigmentation were manually sectioned into thin slices and inspected macroscopically for pigmentation. None was detected. Taken together, these findings suggest that only cells that are developmentally programmed to produce melanin will form pigment when expressing the tyrosinase gene.

Many insights into the development of coat color in mammals have been derived from analysis of chimeric mice generated by aggregating embryos of different genotypes. One of the drawbacks of this approach, however, is the potential for interaction between the genetically different cell types influencing the fate or developmental potential of cells of a given lineage. Single genotype mosaic animals offer a means of analyzing cell lineage and cell differentiation in the embryo in a genetically homogeneous environment. Mosaic mice have been generated previously by retroviral mediated gene transfer to study the allocation of cells of the preimplantation embryo to the somatic and germ cell lineages (Soriano and Jaenisch, 1986), or of cells of the postimplantation embryo to different lineages of the retina (Turner and Cepko, 1987), the central nervous system (Price et al. 1987; Luskin et al. 1988; Walsh and Cepko, 1988; Austin and Cepko, 1990; Galileo et al. 1990) or visceral yolk sac and skin (Sanes el al. 1986). More recently, avian embryos have been infected with recombinant retroviruses to follow the migration of neural crest-derived cells to the gut (Pomeranz et al. 1991) or the dorsal root ganglia (Frank and Sanes, 1991). In this study, we have used a retrovirus vector bearing the human tyrosinase cDNA to mark cells of the neural crest in neurulating albino embryos in order to gain information on the migration and proliferation of progenitor melanoblast clones.

The MLV-HuTyr vector was introduced into the Tnouse embryo by in utero injection. This procedure transduced the tyrosinase gene into the premigratory neural crest population as documented by pigment patches in the coat and eyes of mice derived from injected embryos. The presence of pigmented hairs clearly demonstrates infection of melanoblasts by the vector, and expression of human tyrosinase in the progeny melanocytes. The following considerations argue strongly that each of the pigmented areas was populated by the clonal progeny of a single infected progenitor melanoblast. (1) Only 22% of the injected animals had pigment patches with the majority of the mosaics showing only one single contiguous area of coat pigmentation (29/38 mosaics). This suggests that the overall efficiency of infecting a melanoblast was low and that infection of adjacent melanoblast progenitors would be highly unlikely. (2) The clonal origin of the pigmented areas is also supported by the similar width and shape of the pigment stripes seen in different locations of the trunk in individual mosaics (see below).

Pigmentation in the trunk was typically manifested as transverse stripes of pigment extending ventrally from the dorsal mid-line, reflecting the dorsoventral migration of melanoblast clones. The width of each pigmented band is likely determined by a balance between the proliferative capacity of a marked clone and the population pressure exerted by neighboring uninfected albino clones. In contrast to pigmentation of the trunk, pigment in the head was rarely observed in the form of a stripe, possibly because the shape of the clones was distorted by outward growth of the face, generating the oblique patches observed both in our mosaics and in aggregation chimeras (e.g. see McLaren and Bowman, 1969).

The overall size and shape of retrovirally marked clones in the mosaics are strikingly similar to the coat color patterns of aggregation chimeras, verifying Mintz’s contention that the unit-width bands seen in chimeras represent clonal progeny of a single progenitor melanoblast (Mintz, 1967). Based on the ‘standard’ width of the stripes in aggregation chimeras, Mintz concluded that coat pigmentation was derived from a total of 17 pairs of founder or ‘primordial melanoblasts’ (one for each side of the body), with three pairs contributing to the head, six to the body, and eight to the tail. We have observed considerable variability in the size and shape of pigmented areas in the heads of retrovirally marked mosaics, and have not detected pigment in the tail. The width of marked clones in the trunks of the mosaics, however, is entirely consistent with Mintz’s hypothesis of a small number of clones being allocated to each side of the trunk. Most stripes extended from the mid-dorsal to the mid-ventral lines, i.e. cover the whole territory colonized by a single primordial melanoblast. This suggests that infection of the trunk crest cells occurred at or before allocation to the pigment lineage. In contrast, head crest cells were likely infected after allocation (see below). It is worth noting that in two of the ten mosaics with trunk pigmentation, pigmented hairs extended bilaterally down either side of the animal from a common point of origin on the dorsal mid-line, as if the progeny of a single infected melanoblast were colonizing both sides.

In both mosaics, the stripe on one side extended down almost to the mid-ventral line, but was much shorter and fainter on the other side (compare Fig. 3F). This suggests that some mitotic daughter cells of the marked primordial melanoblasts of one side may have crossed the mid-dorsal line and colonized part of the skin on the contralateral side. Our data raise the possibility that the clonal origins of the left and right sides are not necessarily always independently established, as previously postulated (Mintz, 1970).

Injection of virus on E9.75 produced mosaics with very limited pigmentation, at a frequency 10-fold lower than that obtained on E8.75. Experiments in which marked fibroblasts were microinjected into mid-gestation embryos indicated that on E8.75 the amniotic cavity is the most frequent target of injection; however, on E9.75 marker cells were deposited in the extra-embryonic cavities or decidua rather than the amnion (Huszar et al. 1991). Thus the sharp drop in mosaicism on E9.75 may indicate a requirement for virus to be introduced directly into the amnion where it would have immediate access to the neural crest cells, or neural crest precursors, situated on the closing folds of the neural plate. Both mosaics generated by infection on E9.75 showed faint bands extending only a short distance from the dorsal mid-line. The marking of cells at this stage indicates that melanoblasts are still migrating from the trunk neural tube on E9.75. Similar observations have been made using vital dye labelling (Serbedzija et al. 1990) in which neural crest cells were seen to be continuously migrating from the mouse neural tube between E8.75 and E10.5 along the lateral pathways travelled by melanoblasts. The short length and faintness of the bands may indicate that melanoblasts marked at this stage are subclones of the founder melanoblasts which initiate migration on or before E8.75, and which compete with other, unlabeled, subclones in colonizing an area of the skin occupied by the entire mitotic progeny of the ‘primordial melanoblast.’ Since none of the three subclones marked on E9.75 are found further ventrally along the side of the animal, as patches unconnected to the dorsal mid-line, it appears that melanoblasts are most susceptible to infection on or very near the neural tube, and are much less likely to be accessible while migrating through the mesoderm.

It is of interest to note that the extent of head pigmentation in mosaics infected at E8.75 varied widely, sometimes showing faint patches of only a few pigmented hairs. As appears to be the case for trunk melanoblast clones marked on E9.75, this may also be due to infection of melanoblast subclones rather than founder clones. Neural crest migration in the head is initiated significantly earlier than in the trunk: whereas no crest cell migration from the trunk is observed until embryos develop beyond the 10 somite stage (Erickson and Weston, 1983), the cranial crest begins to migrate from the neural folds at the 4 –6 somite stage, continuing past the 16 somite stage (Nichols, 1981). At the time of injection on E8.75, recipient embryos usually have 10 –15 somites, and it is thus possible that some progenitor melanoblasts in the head have already initiated migration and proliferation, resulting in infection of melanoblast subclones which colonize a reduced area.

A striking correlation was observed between the presence of pigmented areas in the head and in the choroid and iris stroma, suggesting that pigment cells in both locations can be clonally derived from the same melanoblast. In the majority of the mosaics, pigment in the head and eye were located on the same side, however, in about 25 % of these animals the pigmented eye was on the contralateral side of the coat pigment. This may indicate that progeny of a single melanoblast can cross the mid-line and colonize non-contiguous areas. Alternatively, eye pigment in these cases may derive from an independent clone, which contributed only to pigmentation of the eye; in support of this interpretation is the observation that 3 of the total of 38 mosaics generated displayed neural crest-derived eye pigmentation in the absence of any detectable coat pigmentation on the head. In the case of retinal pigmentation, since the pigmented cells of the retinal epithelium derive locally from the optic vesicle, no correlation was seen, as expected, between the presence of neural crest-derived pigment in the coat and CNS-derived pigment in the retina.

Retroviruses injected into postgastrulation embryos can infect cells of most, if not all, tissues and can express the viral genes in all cell types (Jaenisch, 1980; Compere et al. 1989a,b). The tyrosinase gene, under the control of the developmentally unregulated viral LTR, would, therefore, be expected to be active in all infected cell types of the embryo; however, pigment was observed in only subsets of cells, derived from the neural crest and neural epithelium, which are programmed to produce pigment. Bouchard et al. (1989) observed that fibroblasts transfected with the human tyrosinase cDNA occasionally showed pigmented cytoplasmic inclusions, apparently attributable to synthesis and deposition of melanin within cytoplasmic vesicles. We have not, however, detected any pigmented cells in non-melanocyte derivatives of the neural crest or neural epithelium in the mosaics. It is possible that the cytotoxicity of melanin metabolism (Pawelek and Lerner, 1978) eliminates non-pigment cells synthesizing tyrosinase from effectively contributing to development, but we have not seen evidence of overt cytotoxicity in the injected embryos. Alternatively, non-melanoblast-derived cells may produce insufficient amounts of other components required for melanin synthesis.

The experimental approach described in this study establishes the feasibility of transducing genes into neural crest cells in situ, and provides a novel means of experimental access to this pluripotent embryonic cell population. The ability to genetically mark and manipulate neural crest cells in vivo, coupled with the potential for transducing cultured crest cells in vitro and reintroducing them into a developing embryo (Jaenisch, 1985; Huszar et al. 1991) should provide the experimental basis for a systematic investigation of the biology of neural crest cells at the molecular level.

The authors wish to thank Ruth Halaban for helpful advice throughout the course of the study, P. Laird and J. Kreidberg for helpful suggestions and criticism of the manuscript, and Jessica Dausman for competent assistance. D.H. is a recipient of a fellowship from the NCI of Canada, A.S. is a recipient of the Lucille P. Markey fellowship and this work was supported in part by a grant from the Lucille P. Markey Charitable Trust as well as by National Institutes of Health grant (OIG) 5R35-CA44339.