A role for Ultraspiracle, the Drosophila RXR, in morphogenetic furrow movement and photoreceptor cluster formation

The Drosophila eye is composed of an ordered array of 800 unit eyes called ommatidia. These develop from an epithelial monolayer known as the eye imaginal disc. Differentiation of ommatidial clusters occurs in a wave that moves from the posterior margin of the eye field toward the anterior (for review see Thomas and Zipursky, 1994). The margin between the anterior undifferentiated region and the posterior region is marked by an apical-basal constriction of cells, forming a groove in the monolayer known as the morphogenetic furrow (Ready et al., 1976). Cells anterior to the furrow have no set developmental fate. As cells pass into and through the furrow they are organized into 5-cell preclusters (Wolff and Ready, 1991). Upon exiting from the posterior of the furrow, these 5 cells differentiate while others divide and become integrated into the existing preclusters to form the precursors of each individual adult ommatidium (Tomlinson and Ready, 1987).

The propagation of the furrow is dependent on hedgehog (hh), which is expressed posterior to the furrow (Heberlein et al., 1993; Ma et al., 1993). decapentaplegic (dpp), a member of the TGF-β family, is expressed within the furrow and serves as a molecular marker of the morphogenetic furrow (Blackman et al., 1991). Ectopic expression of hedgehog anteriorly leads to the induction of dpp expression and an ectopic furrow. Hh acts by antagonizing the activity of patched (ptc) and cyclic AMP-dependent protein kinase A (Pka-C1). Ptc and Pka-C1 normally repress dpp expression anteriorly (Heberlein et al., 1995; Ma and Moses, 1995; Pan and Rubin, 1995; Strutt et al., 1995) and mutations in ptc and Pka-C1 lead to ectopic anterior expression of dpp.

The anterior movement of the morphogenetic furrow and neuronal differentiation is also regulated by four helix-loop-helix (HLH) proteins, Hairy (H), Extramacrochaetae (Emc), Atonal (Ato) and Daughterless (Da). Clones of cells mutant for either h or emc alone show no or very subtle defects in eye development, but double mutant clones show advanced furrow movement and premature neuronal differentiation (Brown et al., 1995). Ectopic anterior expression of hh, in addition to inducing a new furrow and dpp expression, leads to Hairy expression ahead of the new furrow (Heberlein et al., 1995). atonal and da expression are markers of, and are possibly causal of, the first neuronal differentiation seen in the eye disc. ato expression begins just anterior to dpp expression and appears to be in every cell. da is expressed anterior of the morphogenetic furrow and shows higher expression within the furrow. As the furrow passes, ato and da expression become limited to the cells that will become the R8 photoreceptors and thereby establish the position and spacing of ommatidial clusters (Brown et al., 1995; Jarman et al., 1994).

Ultraspiracle (Usp) is the Drosophila cognate of RXR (Oro et al., 1990). Like RXR, it forms heterodimers with vertebrate receptors including RAR, the Thyroid Hormone Receptor and the Vitamin D Receptor, as well as with the Drosophila Ecdysone Receptor and DHR38 (Sutherland et al., 1995; Yao et al., 1993). The complex phenotype of usp suggests roles in several signaling pathways (Oro et al., 1992). We have previously demonstrated that usp mutations alter aspects of terminal differentiation in the Drosophila eye. The phenotypes include abnormal rhabdomere development as well as loss of expression of Rh-4, an R7 photoreceptor specific opsin (Oro et al., 1992; Yao, 1994 and unpublished results). In the course of studying these phenotypes, we examined the earliest stages of differentiation in third instar eye imaginal discs. We show here that usp is required for proper progression of the morphogenetic furrow and subsequent organization of the ommatidial clusters. Our data suggest that usp defines a third pathway for regulation of furrow movement separate from the hairy/emc and patched/Pka-C1 pathways. We propose that usp represses the propensity of anterior cells to enter into the ‘furrow fate’ in response to signals coming from near or posterior to the morphogenetic furrow and that loss of usp function allows precocious differentiation at the anterior margin of the furrow.

The generation and detection of mitotic clones in imaginal discs was carried out using the FLP/FRT/myc system as described (Xu and Rubin, 1993). Enhancer trap or reporter lines were first crossed to either w usp³f FRT18A/w usp³f FRT18A; λ10 Tb/TM3 or y w usp⁴ FRT18A/y w usp⁴ FRT18A; λ10 Tb/TM3. λ10 indicates an 8 kb usp⁺ genomic transgene inserted on the third chromosome (Oro et al., 1992). Males that contained the reporter and were of the general genotype usp FRT18A; λ10 Tb were crossed to π-myc M(1)o^Sp FRT18A/FM7; hsFLP38/hsFLP38 or N-myc M(1)o^Sp FRT18A/FM7; hsFLP38/hsFLP38. For the emc h experiments, a dpp 3.0/dpp 3.0; emc¹h^c1 FRT80/ TM6B stock (a gift from Nadean Brown) was crossed to hsFLP22/hsFLP22; π-myc FRT80/π-myc FRT80. In experiments with both usp and emc h, first and second instar larvae were subjected to a 1 hour heat shock at 37°C to induce mitotic recombination via expression of the Flipase. usp³ and usp⁴ are mis-sense mutations in the DNA-binding domain which affect DNA binding in conjunction with the Ecdysone Receptor. They are inferred to be strong hypomorphic alleles (Henrich et al., 1994; Oro et al., 1992) (and unpublished results). usp³ and usp⁴ mutant clones give identical phenotypes in the eye disc. pGMR-usp was constructed by inserting a usp cDNA into the EcoRI site of pGMR (Hay et al., 1994) and introduced into flies as described (Boggs et al., 1987). Similar results are observed using either of two pGMR-usp insertions into the third chromosome.

Fixing and staining procedures have been described (Xu and Rubin, 1993). Primary antibodies: mouse anti-Usp diluted 1:5 (gift from F. Kafatos), rabbit anti-lacZ (Jackson Labs) diluted 1:1500, rabbit antiAtonal (gift from Y. N. Jan) diluted 1:1000, mouse anti-myc (Oncogene Science) diluted 1:100, rabbit anti-hairy (gift from S. Carroll) diluted 1:200, rat anti-Elav (Developmental Studies Hybridoma Bank) and rabbit anti-cyclin A diluted 1:1000 (gift from D. Glover). Secondary antibodies: Texas Red-conjugated goat antimouse and FITC-conjugated goat anti-rabbit and goat anti-rat (Jackson Labs) diluted 1:100. Confocal images were collected on a Nikon/Biorad confocal microscope and imported into Adobe Photoshop for presentation.

Usp is expressed in all cells of the eye imaginal disc (Fig. 1A). The apparent higher expression just anterior to the morphogenetic furrow is due to the intrinsic physical characteristics of the eye imaginal disc. Similar nuclear staining is seen when assayed by the DNA stain DAPI (data not shown). To determine the role of usp in the development and differentiation of the Drosophila eye, we created clones of cells that are mutant for usp using the FLP/FRT system (Xu and Rubin, 1993) and examined the consequences using expression of individual genes or proteins as markers of events related to eye differentiation. All phenotypes observed are rescued by the presence of an 8 kb wild-type usp transgene (Fig. 2G-I).

In usp⁻ clones that lie posterior to the morphogenetic furrow, we see abnormal arrangement of developing photoreceptor clusters (Fig. 2A-F). Photoreceptor clusters are irregularly spaced, resulting in gaps between the clusters. Similar results are seen with multiple markers of cell fate (Fig. 2, data not shown). With each marker used, including an antibody against ELAV (Bier et al., 1988), which marks each of the eight photoreceptors, we see that clusters still contain the appropriate number of cells (data not shown).

Not only are ommatidial clusters misaligned in mutant regions, mutant cells posterior to the furrow differentiate prematurely (Fig. 2A-C). Using the spalt-expressing enhancer trap line 2-3602 (our unpublished results), we see labeled clusters of five cells starting five to six rows posterior to the furrow in wildtype regions, whereas in the usp⁻ clones they appear three rows from the furrow. Even though 5-cell clusters appear prematurely in the usp⁻ regions, the number of 2-3602-expressing cells and their relative arrangement within the clusters appears normal.

When usp function is absent, morphogenetic furrow movement is accelerated, resulting in a furrow, which bends anteriorly within and around the usp⁻ region (Figs 2D-F, 3D-F). The furrow can become advanced by as many as three to four rows of ommatidial clusters. Bending is observed using markers of differentiation (Fig. 2D-F) or markers that are expressed within the furrow (Fig. 3D-F). The effect is progressive and thus dependent on the shape and size of the clone. As observed in Fig. 3A-C, when the furrow is just entering the posterior region of the clone, there is no anterior bending of the furrow. As the morphogenetic furrow passes through a large posterior-anterior spanning clone, the bending anteriorly is more prominent. Slight furrow displacement continues even after the furrow has re-entered the wild-type region (Fig. 2A-C, data not shown).

The anterior bending of the furrow might result from an excess of cells in the usp⁻ clones. For example, the premature differentiation seen in usp⁻ clones may include an acceleration of the final cell divisions that occur posterior of the furrow (Ready et al., 1976). Thus the bending of the furrow would not be caused by loss of usp anterior to or within the furrow region but rather by a mechanical pushing of the furrow as a result of an increase in the number of cells in mutant regions posterior to the furrow. There are several reasons to discard this hypothesis.

A ‘pushing effect’ from extra cell proliferation might be expected to lead to expansion and deformation in all directions, thus creating a non-autonomous disruption of the pattern of cells in all directions outside of the clonal boundary. We would expect to see not only an anterior displacement of the furrow but also a displacement of wild-type ommatidial clusters lying dorsally, ventrally and posterior of the usp⁻ clones. We have not seen this effect in any usp⁻ clones. The disruption of the ommatidial pattern ends at the clonal boundary.

To examine directly the possibility of increased cell density, we used DAPI stain to mark nuclei of both wild-type and mutant cells. In clones anterior to, intersecting or posterior to the morphogenetic furrow, we see no increase in cell density (Fig. 4A-C). Thus, neither deformation of adjacent lateral regions nor an increase in cell density are seen in or around usp⁻ clones.

We also characterized the density of actively cycling cells by using the expression of cyclin A as a marker of cells in the S and G2 phases of the cell cycle. Cyclin A expression is normal in clones that intersect the morphogenetic furrow (Fig. 4D-F and data not shown). In usp⁻ clones that lie posterior to the morphogenetic furrow, there appear to be fewer cyclin A-positive cells than in adjacent wild-type regions (Fig. 4D-F). We interpret this to be a indication that there are fewer, not more, cycling cells in older usp⁻ regions posterior to the morphogenetic furrow.

The disordered arrays of clusters and altered furrow advancement are similar to phenotypes characterized for other pathways that regulate furrow movement and the early steps of differentiation and may implicate Usp as a yet unidentified member of these signaling pathways. Given the ability of RXR-containing dimers, including Usp-containing dimers, to repress transcription (Chen and Evans, 1995; Horlein et al., 1995 and unpublished data), it is possible that Usp may be involved in the repression of anterior differentiation by Ptc and Pka-C1. To test this hypothesis, we asked if dpp expression, a marker of Ptc and Pka-C1 activity, occurs anterior to the furrow in usp⁻ clones (Heberlein et al., 1995; Pan and Rubin, 1995; Strutt et al., 1995). Ptc and Pka-C1 normally repress dpp expression anteriorly and mutations in ptc and Pka-C1 lead to ectopic anterior expression of dpp. If Usp protein acts downstream of Ptc and Pka-C1 to repress gene expression, or if Usp acts upstream of ptc and Pka-C1 to allow their expression, then loss of usp function should lead to an increase in dpp expression anterior to the furrow as is observed in ptc⁻ and Pka-C1⁻ clones.

As can be seen, in usp⁻ clones there is no anterior expansion or ectopic expression of dpp (Fig. 3A-F and data not shown) and dpp expression continues to follow the furrow as it bends anteriorly within usp⁻ clones (Fig. 3D-F). The deformed morphogenetic furrow also encompasses some adjacent usp⁺ cells, similar to what is seen in Fig. 2D-F, almost certainly as a result of the non-cell autonomous action of Hh in induction of differentiation. We have also examined expression of an hh enhancer trap in usp⁻ clones. As can be seen in Fig. 5, the appropriate relationship between hh expression and the morphogenetic furrow is seen, with hh expression beginning posterior to the furrow. This is true even in situations in which the furrow is anteriorly displaced. These results demonstrate that usp is not required for ptc and Pka-C1 expression and that ptc and Pka-C1 functions do not involve usp activity to repress dpp expression. This strongly suggests that usp and ptc and Pka-C1 define independent pathways controlling furrow movement.

hairy (h) and extramacrochaetae (emc), two genes encoding helix-loop-helix proteins represent a second set of genes regulating anterior movement of the morphogenetic furrow. To test whether usp is required for hairy transcription and thus is a molecule involved in this signaling pathway, we looked at Hairy expression in usp⁻ clones. As can be seen, Hairy is normally expressed in a stripe just anterior to the region in which atonal is expressed. This pattern is unchanged in usp⁻ cells (Fig. 6A-C). Since the hairy emc mutant phenotype is only seen when both hairy and emc are mutant, the normal expression of Hairy in a usp mutant clone indicates that the usp phenotype is not a consequence of the loss of hairy and emc expression. In addition, usp is expressed normally in hairy⁻emc⁻ cells (data not shown). These data are consistent with a model in which usp and hairy/emc are part of separate pathways for the control of furrow movement.

As a test of the hypothesis that usp and hairy are in different pathways, we examined the effect of usp mutations on one likely target of down regulation by hairy, atonal (Brown et al., 1995; Jarman et al., 1994). In addition, analysis of the initial postfurrow pattern of Ato expression in usp clones allowed us to determine if the disarrayed ommatidial clusters seen in usp⁻ clones result from gross errors in the initial patterning of the R8 photoreceptor precursors. To test these possibilities, we examined the expression of Ato in usp⁻ clones. In Fig. 7A-C, a usp⁻ clone intersects the normal Atonal-expressing region just anterior to the morphogenetic furrow with no obvious anterior expansion of Atonal expression. In addition, the limitation of Atonal expression to the presumptive R8 photoreceptor cells in the region of the morphogenetic furrow occurs with apparently normal kinetics in usp⁻ clones (Fig. 7D-F). Examination of Ato expression patterns in usp mutant regions well posterior to the furrow indicates that the array and spacing of R8 cells is normal (Fig. 7G-I), even though the apparent ommatidial arrays are abnormal. Thus, Usp does regulate the kinetics or position of atonal expression, and the loss of Usp function does not alter the patterning of the R8 precursors.

The data presented indicate that usp acts in a pathway independent of both ptc/Pka-C1 and hairy/emc. Small differences in phenotype between ptc or Pka-C1 clones and hairy emc clones suggest that ptc/Pka-C1 and hairy/emc define separate pathways for control of furrow movement (Brown et al., 1995; Heberlein et al., 1995; Ma and Moses, 1995; Pan and Rubin, 1995; Strutt et al., 1995). In order to test this directly, we again used dpp as a marker of the function of the ptc/Pka-C1 pathway. If hairy and emc function in the same ptc/Pka-C1 pathway that regulates dpp, then loss of hairy and emc should lead to ectopic expression of dpp in the region anterior to the morphogenetic furrow, just as occurs in ptc or Pka-C1 mutant clones. As can be seen, hairy emc mutant clones, which intersect the furrow from the posterior, have an anteriorly shifted furrow with dpp expression in the furrow (Fig. 8A-C). In a clone that crosses the furrow from posterior to anterior, there may be the beginning of furrow advancement or a small anterior expansion of dpp expression, but dpp expression does not fill the anterior region of the clone as would have happened in a ptc or Pka-C1 clone (Fig. 8D-F). Clones that lie entirely anterior to the furrow do not show dpp expression (data not shown). These phenotypes are different from those of ptc or Pka-C1 mutant clones. Thus, we conclude that ptc/Pka-C1, hairy/emc and usp each define a separate regulatory pathway acting to control events related to the rate of morphogenetic furrow movement.

From the data presented, it is not clear whether usp functions in the posterior differentiated region, in the anterior undifferentiated region, or both. To distinguish among these possibilities, we have constructed a transgene, pGMR-usp, which leads to high level usp expression posterior to the furrow as verified by anti-Usp antibody staining (Fig. 9A) (Hay et al., 1994). Examination of usp⁻ clones in a pGMR-usp background reveals that photoreceptor clusters within the usp⁻ clone are irregularly organized, even though Usp is being expressed at high levels from pGMR-usp (Fig. 9B-G). Some residual anterior deformation of the furrow can be seen (Fig. 9E-G). These results indicate that the levels or position of Usp expression from pGMR-usp are insufficient to rescue the usp mutant phenotype. Since pGMR-usp gives relatively high Usp expression in most regions posterior to the furrow (Fig. 9A), the simplest interpretation of these results is that usp normally functions anterior to, within or just posterior to the morphogenetic furrow.

Previous results from our laboratories demonstrate a requirement for Usp, the Drosophila RXR, in the development of the adult compound eye (Oro et al., 1992). The results in this paper demonstrate that Usp plays a role from the earliest stages of cell determination and differentiation. Regions of the eye imaginal disc that lack usp show more rapid morphogenetic furrow movement, earlier post-furrow differentiation, as judged by expression of cell-type-specific markers, and abnormal ommatidial cluster formation and alignment.

Although a usp⁺ genomic fragment rescues the usp phenotypes, these phenotypes are not rescued by above normal expression of Usp posterior to the morphogenetic furrow. This leads us to conclude that the array of usp mutant phenotypes observed in the early stages of eye differentiation result from loss of usp activity in the region anterior to or within the morphogenetic furrow.

From the observation that events proceed quicker in the absence of usp function, we infer that Usp, directly or indirectly, inhibits or slows the rate of furrow movement and differentiation. A direct inhibitory action of Usp, perhaps in conjunction with a non-Usp dimerization partner, is consistent with our observations that Usp/EcR dimers can repress basal level transcription (unpublished observations) and with experiments demonstrating a repressive function for unliganded RXR and some of its dimerization partners (Chen and Evans, 1995; Horlein et al., 1995).

In addition to usp, two other sets of genes have been shown to slow the rate of morphogenetic furrow movement. These are ptc/Pka-C1 (Heberlein et al., 1995; Ma and Moses, 1995; Pan and Rubin, 1995; Strutt et al., 1995) and hairy/emc (Brown et al., 1995). This raises the question as to whether these define separate regulatory pathways or whether they all act in a single, linear cascade. Our results, when coupled to the results of others, show that ptc/Pka-C1, hairy/emc and usp define separate pathways inhibiting furrow movement.

First, we show that neither hairy/emc nor usp mutant regions in the anterior portion of the eye disc show complete ectopic expression of dpp as is observed in regions mutant for ptc or Pka-C1. Thus, hairy/emc and usp can be separated from the ptc/Pka-C1 pathway. Second, our results show that hairy is not required for Usp expression and usp is not required for Hairy expression. Finally, our results show that atonal, a target for repression by hairy/emc, is expressed normally, relative to the morphogenetic furrow, in usp mutant regions, even though misexpression of Atonal protein occurs in hairy/emc mutant regions (Brown et al., 1995). Thus we conclude that not only are the hairy/emc and usp pathways separate from the ptc/Pka-C1 pathway, they are separate from each other.

That usp defines a separate pathway controlling furrow movement does not explain the mechanism by which it acts. The progressive nature of furrow bending in usp clones suggests that small increments of furrow advancement are compounded as the furrow moves through the usp mutant regions. These incremental advancements could broadly reflect two possibilities. (1) The loss of usp function could result in increased sensitivity of anterior cells to a signal generated in the differentiating posterior region or very near the furrow (Strutt et al., 1995). Under such a model, with both the ptc/Pka-C1 and h/emc pathways still functioning, we can infer that usp must normally function relatively close to the morphogenetic furrow in repressing the response to such a differentiation factor. (2) As a consequence of more rapid differentiation of usp⁻ cells just posterior to the furrow, higher levels of a differentiation-inducing factor may be generated, leading to earlier commitment to the furrow fate and anterior bending of the furrow. Under this model, hedgehog protein is the obvious candidate for such a molecule. In usp mutant clones, the relationship between the furrow and the hedgehog-expressing cells is maintained and the relative level of hedgehog expression is normal. In addition, our studies show that the earliest stages of Atonal expression within and posterior to the furrow are normal. Given that two different markers of the earliest stages of postfurrow differentiation appear normal, we suggest that it is more likely that usp modulates the response to a signal in the anterior or furrow region rather than modulating the generation of a signal in the posterior region.

Within usp⁻ clones, ommatidial clusters are misaligned with respect to each other and with respect to adjacent wild-type tissue. Our experiments suggest that this phenotype results from loss of usp function anterior to, within or just posterior to the morphogenetic furrow. This suggests that usp may function in the earliest stages of ommatidial cluster formation. The initial determination of the number, positioning and spacing of ommatidial clusters is the designation of single cells as the precursors of the R8 photoreceptor cells of each individual ommatidium. Precluster formation then occurs around these R8 precursor cells. Using Atonal expression as a marker of R8 determination and position, we see normal kinetics for R8 specification with the same density of R8 precursors as seen in nearby wild-type regions. In older mutant regions of the disc, the Atonal-expressing cells fall in a normal array comparable to that seen in adjacent wild-type regions, even though other markers of differentiation show that clusters are abnormal in their alignment to each other.

One possibility is that the disruption of patterning is due to the premature or abnormal differentiation and recruitment of cells into ommatidial clusters that is seen in mutant clones. The exact nature of the premature differentiation events is not clear. The pattern of expression of the 2-3602 enhancer trap is consistent with it being a marker of R1, R3, R4, R6 and R7 differentiation. If so, the pattern of expression in usp⁻ clones suggests that R1, R6 and R7 enter ommatidial clusters prematurely. Alternatively, the ‘mystery cells’ (Wolff and Ready, 1993) that are normally eliminated from ommatidial preclusters may remain and differentiate as 2-3062-expressing cells. If this is the case, these modified mystery cells must take on the relative positions of the normal 2-3602-expressing cells and inhibit recruitment of the cells otherwise destined to become the 2-3602-expressing cells. In either case, based on the DAPI and cyclin A staining and premature appearance of molecular markers, it is possible that cells taking on photoreceptor fates, and possibly some of the non-neuronal accessory cells, may be recruited into or retained within the preclusters before the synchronous round of cell division that occurs posterior of the furrow (Ready et al., 1976). This could result in too few uncommitted cells being present at later stages for adding the final non-neuronal cells to clusters, thus resulting in the poor relative organization of the mutant ommatidial clusters.

We conclude that usp, the Drosophila RXR, defines a novel pathway for control of morphogenetic furrow movement and differentiation. That usp may function anterior to the furrow to modulate the response to a factor arising posterior to or very near the morphogenetic furrow raises a number of questions including what signaling pathway involves usp and what, if anything, does usp tell us about the role of RXRs in vertebrate limbs? Results from Drosophila and the high degree of conservation between the Drosophila eye and vertebrate limbs lead us to suggest a possible answer to these questions.

Others have inferred the existence of a non Dpp ‘competency’ factor functioning anterior to the morphogenetic furrow (Burke and Basler, 1996; Strutt et al., 1995). Work on vertebrates has identified a positive feedback loop between sonic hedgehog (shh) expression in the zone of polarizing activity and fibroblast growth factor (FGF) expression in the Apical Ectodermal Ridge such that Shh induces FGF and FGF induces Shh (Laufer et al., 1994; Niswander et al., 1994; Vogel et al., 1996; for a review see Cohn and Tickle, 1996). Recent studies have identified both a Drosophila FGF and FGF receptors, with one FGF receptor, heartless, being expressed in a region at or near the morphogenetic furrow (Emori and Saigo, 1993). We therefore propose the possibility that a Drosophila FGF, acting via heartless, may be the ‘competency’ factor identified by others and that usp may function to down regulate the response to FGF in regions anterior to the morphogenetic furrow. If so, this may suggest a similar role for RXR in controlling the boundary between FGF- and Shh-expressing cells in vertebrate limb buds.

In conclusion, we have shown the involvement of an RXR family member in regulating the anterior/posterior boundary, the morphogenetic furrow, in the developing Drosophila eye and for regulation of posterior differentiation and ommatidial cluster formation. We have also shown that the function of Usp in these events defines a previously unrecognized pathway for the control of furrow movement and cell differentiation. Given the high degree of similarity between Drosophila and vertebrate regulatory pathways, these results have important implications for future studies of vertebrate eye and limb development.

We thank Drs Y. Hiromi, Y. N. Jan, S. B. Carroll, F. Kafatos, D. Glover, N. Brown and the Bloomington Drosophila Stock Center for antibodies and fly stocks. We thank P. Edeen for technical support and T.-P. Yao for helpful discussions. We are grateful to the UCSD Bioengineering Department for use of the confocal microscope and the advice of Dr J. Price. This work was supported by a grant from American Cancer Society to M. M.; A. C. Z and N. G. were supported in part by a predoctoral training grant from the National Institute of Child Health and Human Development. A. C. Z. is a Chapman Fellow of The Salk Institute. N. G. is a Fellow of the Salk Institute Associates. C. T. is a Damon Runyon-Walter Winchell postdoctoral fellow. M. M. and R. M. E. are members of the Cancer Center at the Salk Institute. R. M. E. is an investigator of the Howard Hughes Medical Institute at The Salk Institute.