Genetic and molecular characterization of the dominant suppressors of D-rafC110 on the second chromosome identified two gain-of-function alleles of rolled (rl), which encodes a mitogen-activated protein (MAP) kinase in Drosophila. One of the alleles, rlSu23, was found to bear the same molecular lesion as rlSem, which has been reported to be dominant female sterile. However, rlSu23 and the current stock of rlSem showed only a weak dominant female sterility. Detailed analyses of the rl mutations demonstrated moderate dominant activities of these alleles in the Torso (Tor) signaling pathway, which explains the weak dominant female sterility observed in this study. The dominant rl mutations failed to suppress the terminal class maternal-effect mutations, suggesting that activation of Rl is essential, but not sufficient, for Tor signaling. Involvement of rl in cell proliferation was also demonstrated by clonal analysis. Branching and integration of signals in the MAP kinase cascade is discussed.
MITOGEN-activated protein kinase (MAP kinase or MAPK) plays essential roles in the transduction of diverse extracellular signals regulating cellular proliferation and differentiation. Its activity is closely regulated by phosphorylation of both threonine and tyrosine residues in its activation loop by a dual-specificity kinase MAP kinase kinase (MAPKK or MEK). MAPKK is also regulated by phosphorylation of two adjacent serine/threonine residues in its activation loop by MAP kinase kinase kinase (MAPKKK or MEKK). This cascade of protein kinases, known as the MAPK cascade, is highly conserved during evolution and found ubiquitously among eukaryotes (Nishida and Gotoh 1993; Davis 1994).
Drosophila also contains the MAPK cascade, and rolled (rl) (Biggs and Zipursky 1992; Biggset al. 1994; Brunneret al. 1994), Dsor1 (Tsudaet al. 1993; Luet al. 1994), and D-raf (Nishidaet al. 1988; Ambrosioet al. 1989) have been identified as encoding the components of the cascade, MAPK, MAPKK, and MAPKKK, respectively. Genetic analyses revealed the involvement of these genes in the transduction of signals from the receptor tyrosine kinases (RTKs) encoded by sevenless (sev), torso (tor), and Drosophila EGF receptor homolog (DER; Ambrosioet al. 1989; Dickson et al. 1992, 1996; Tsudaet al. 1993; Biggset al. 1994; Brand and Perrimon 1994; Brunneret al. 1994; Hataet al. 1994; Hsu and Perrimon 1994; Nishidaet al. 1996; Limet al. 1997). The RTKs encoded by sev and tor are responsible for the determination of the R7 photoreceptor cell fate in the eye disc and of the cell fates at the embryonic termini, respectively (Hafenet al. 1993; Duffy and Perrimon 1994). DER has multiple functions, such as the determination of dorso-ventral polarity of the ovarian follicle, the embryonic ectodermal differentiation, and the differentiation of the compound eye and wing veins (Shilo and Raz 1991). In addition, D-raf and Dsor1 have been demonstrated to be involved in the regulation of cellular proliferation (Perrimonet al. 1985; Nishida et al. 1988, 1996; Tsudaet al. 1993; Hataet al. 1994). Thus, the ubiquitous MAPK cascade receives diverse extracellular signals and generates responses specific to each RTK. Elucidation of the molecular mechanism by which signal specificity is generated will be crucial for understanding the molecular mechanisms of development.
It has been demonstrated that the transduction of the signals generated by different RTKs is mediated through a cassette of pathways composed of multifunctional factors encoded by drk, Sos, Ras1, D-raf, Dsor1, and rl (Ambrosioet al. 1989; Doyle and Bishop 1993; Luet al. 1993a; Tsudaet al. 1993; Brunneret al. 1994; Diaz-Benjumea and Hafen 1994). However, the cassette does not seem to be a simple, straightforward pathway, and the signal specificity in each RTK pathway may be provided by branching and integration of signals as well as the differential modulation of some components. For example, it has been proposed that a Ras1-independent pathway activates D-raf in the Tor pathway from the observation that a loss of Ras1 activity did not completely abolish Tor signaling (Houet al. 1995). Genetic analysis of gain-of-function mutations of Dsor1 demonstrated that Dsor1 transduces signals less efficiently in the DER pathway than in the Tor or Sev pathways, suggesting that differential modulation of the Dsor1 activity may be involved in generating signal specificity (Limet al. 1997). We herein characterize gain-of-function mutations of rl and their activities in Tor signaling and in imaginal cell proliferation. The results obtained in this study suggest novel points for the integration or branching of signals in the MAPK cascade.
MATERIALS AND METHODS
Genetics: Fly cultures and crosses were performed at 25° unless otherwise described. Fly stocks used in this study were provided as follows: fs(1)ph1901/FM6 and torRL3 cn px sp/CyO from G. Struhl; D-rafC110; rlSem from D. Brunner and E. Hafen; w; l(2)rlEMS64 Pin/SM1 from D. Yamamoto; and rl1, Df(2R)rl10a, lt rl10a cn/SM1 and Df(2R) rl10b, lt rl10b cn/SM1 from the Bloomington Stock Center. For descriptions of the genetic markers and balancers, see Lindsley and Zimm (1992). D-rafC110/Y; rlSem/rlSem males were crossed with Canton-S (wild-type) females and their male progeny were further crossed with Canton-S females to remove D-rafC110. Females heterozygous for rlSem (+/+;rlSem/+) were selected by the dominant eye/wing phenotypes and tested for dominant sterility.
Clonal or twin-spot analysis was performed with Dp(1; Y; 3)M2′, mwh+ ve+ FR1, y cv v f36a/C(1)RX, y f/BSY; mwh ve h as described earlier (Lawrenceet al. 1986; Tsudaet al. 1993). Females with relevant mutations that were also homozygous for mwh were crossed with males of the above strain, and their progeny were irradiated with X ray (1500 R) during early third instar (∼72 hr after eggs were laid). Clones formed in wing blades were analyzed, and only those twin-spots with >16 f36a (control) cells were considered.
Molecular procedures: RNA was extracted from homogenized adult flies homozygous for either rlSu14 or rlSu23, and cDNAs were synthesized using oligo(dT) primers and Superscript reverse transcriptase (Boehringer, Indianapolis). The mutant rl cDNAs were cloned as two overlapping fragments synthesized by RT-PCR using sets of sense and antisense primers synthesized according to the rl coding sequence (Biggs and Zipursky 1992). The sense primers were 5′-GAGGATTCCGACAAGTGAATTTATTCTATTTCACCC-3′ and 5′-GAGGATTCCATGTCACAAACTACCTCAGAC-3′, and the antisense primers were 5′-GAGGATCCTCCCGATGCAAGACGT TTGCGGAATG-3′ and 5′-GAGGATTCGCAAAATGGAGAAG TCCAGC-3′. Each primer contained a BamHI restriction site (underlined) to facilitate ligation into the BamHI cloning site in pBluescript (Stratagene, La Jolla, CA). Three independent clones were sequenced on each strand using synthetic oligonucleotide primers.
Whole-mount in situ hybridizations: Digoxigenin-labeled antisense and sense RNA probes were made from linearized plasmid DNAs containing tll, hkb, or ftz cDNA fragments using the DIG RNA labeling kit (Boehringer). Probes were treated with alkali to reduce their sizes to an average of 100 nucleotides long and were used for in situ hybridization with whole-mount embryos following the method of Tautz and Pfeifle (1989), with slight modifications.
Other procedures: Histological sections for electron microscopy were prepared as previously described (Basleret al. 1991). Cuticle preparations of embryos were made as described by Wieschaus and Nüsslein-Volhard (1986), and viewed with dark-field optics.
Genetic and molecular characterization of gain-of-function mutations of rolled: To identify factors acting downstream of D-raf, we screened for dominant suppressors of a hypomorphic allele of D-raf, D-rafC110, and obtained 19 such mutants (Tsudaet al. 1993; Limet al. 1997). In this article, the two second chromosomal suppressors, Su14 and Su23, are described. These mutations fully suppressed the phenotypes of D-rafC110 in terms of the viability and the morphologies of the compound eye and wing veins (data not shown). Both mutations mapped genetically at 55.0 in the centromeric heterochromatic region of the second chromosome, where rolled (rl), encoding a Drosophila homolog of MAPK, is located (Biggset al. 1994).
In the D-raf+ background, Su23 caused a dominant phenotype, producing a mild rough eye and extra wing veins (Figure 1B and Figure 2B). Observation of ultrathin sections of the compound eye revealed multiple R7-like cells in each ommatidium (Figure 1D). Flies heterozygous for Su14 showed no apparent rough eye phenotype, but observations of their eye sections revealed extra R7-like cells in a small fraction of ommatidia. Flies homozygous for Su14 showed a mild rough eye phenotype with multiple R7-like cells in most of the ommatidia (Figure 1E). Both mutations strongly suppressed the loss-of-function mutation of sev, and extra R7-like cells were produced in Su23 even with the genetic background of a null sev mutation (data not shown).
A gain-of-function mutation of rl, rlSevenmaker (rlSem) resulted in a similar phenotype that was significantly enhanced by loss-of-function mutations of rl (Brunneret al. 1994). We observed that rl mutations significantly enhanced the wing phenotype seen in Su23, and that more wing vein materials were produced in Su23/rl1 than in Su23/+ (Figure 2D). The eye phenotype was also enhanced and more R7-like cells were produced in Su23/rl1 and Su23/Df(2R)rl10b flies [3.19 ± 1.30 (N = 97) and 3.18 ± 1.23 (N = 90) R7-like cells per ommatidium, respectively; Figure 1F] than in Su23/+ flies (2.75 ± 1.33 R7-like cells per ommatidium, N = 122). Although Su14/+ showed no extra wing veins, flies transheterozygous for rl1 and Su14 did (Figure 2C). The enhancement of the dominant activities may be due to lack of competition from the normal product, and the results may suggest that they are the alleles of rl.
To confirm the allelism further, we cloned the rl cDNA fragments by RT-PCR with template RNAs extracted from flies homozygous for the suppressor mutations, and sequenced them as described in materials and methods. Comparison of the nucleotide sequences with those reported (Biggs and Zipursky 1992) revealed guanine-to-adenine substitutions at nucleotide position 737 in Su14 and 1214 in Su23, respectively, causing changes from aspartic acid to asparagine at amino acid residue 185 in kinase subdomain VII in Su14 and at 334 in kinase subdomain XI in Su23, respectively (Figure 3). Both residues are highly conserved among MAPK family members, and the alterations are likely the cause of the suppressor mutations. Hereafter, we refer to them as rlSu14 and rlSu23. To our surprise, the molecular lesion associated with rlSu23 was identical to that found in rlSem (Brunneret al. 1994).
Activity of rolled in the Torso pathway: It has been reported that flies heterozygous for rlSem are almost invariably dominant female sterile (Brunneret al. 1994). On the contrary, rlSu23/+ females are fertile, and a detailed analysis revealed a weak dominant sterility. A fraction of the eggs (10.9%, N = 2688) laid by rlSu23/+ females crossed with normal males failed to hatch, and observation of the cuticular pattern of the nonviable embryos revealed that 15.6% (N = 205) of them were associated with reduction in the number of abdominal segments (data not shown). This phenotype is similar to that observed in embryos produced by females carrying gain-of-function muations in tor (Klingleret al. 1988). The remaining nonviable embryos showed no obvious aberration. Reexamination of rlSem, kindly provided by D. Brunner and E. Hafen, also demonstrated a weak dominant sterility similar to rlSu23. The rlSem/+ flies produced nonviable embryos (10.2%, N = 2066), a fraction of which (20.0%, N = 184) showed defects similar to those associated with the gain-of-function alleles of tor. Thus, both rlSu23 and the current stock of rlSem are only weakly dominant female sterile. Females homozygous for rlSu23 are essentially sterile due to defects in vitellogenesis (data not shown). Due to this discrepancy, we analyzed the effects of the rl mutations in the terminal system in more detail.
The effects of the gain-of-function mutations of rl on the expression of tll and hkb, target genes in the Tor signaling pathway (Pignoniet al. 1990; Weigelet al. 1990; Tsudaet al. 1993), were examined. In normal blastoderm embryos, tll is expressed in a pattern having a posterior cap and an anterior dorsal-lateral stripe (Pignoniet al. 1990; Tsudaet al. 1993; Figure 4A). In the embryos derived from the tor dominant females, the tll expression region is greatly expanded into the central region, and the expression pattern of the pair-rule gene fushitarazu (ftz) is severely affected so as to produce only three stripes instead of the normal seven stripes (Steingrimssonet al. 1991; Streckeret al. 1991). In the embryos laid by rlSu23/+ or rlSu14/+ females (Figure 4, B and C), no gross alteration of the tll expression pattern was observed, although the tll-expressing regions were significantly expanded. The anterior border of the posterior tll expression region was shifted to 19.3% egg length (EL) and 18.8% EL in the embryos laid by rlSu23/+ or rlSu14/+ females, respectively, as compared to 14.6% EL in normal embryos. There was also a significant posterior shift in the anterior border of the anterior stripe in these embryos (Figure 4, B and C). Another target gene, hkb, is expressed at both termini in blastoderm embryos (Weigelet al. 1990; Figure 4F), and its expression was also significantly expanded at both termini in these embryos (Figure 4, G and H). No gross alteration of the expression pattern of ftz was observed, although the seven stripes were considerably condensed toward the central region (Figure 4L). As most of the embryos develop normally, the altered pattern of development would thus be regulated during later stages. Similar results were obtained with the embryos laid by rlSem/+ females (data not shown).
We also analyzed the genetic interactions of rl with the terminal class maternal-effect mutations. Embryos produced by females homozygous for terminal class maternal-effect mutations such as fs(1)ph fail to develop structures posterior to the seventh abdominal segment as well as anterior-most structures, including the head skeleton (Figure 4M and Table 1). Expression of tll and hkb is severely affected in these embryos (Weigelet al. 1990; Tsudaet al. 1993; not shown, but essentially identical to expression shown in Figure 4, E and J). Dsor1 has been demonstrated to act downstream of D-raf in the Tor pathway, and the gain-of-function mutations of Dsor1 suppress the posterior defects in embryos devoid of terminal class gene maternal activities (Tsudaet al. 1993; Limet al. 1997; Table 1; Figure 4, D, I, and N). The loss-of-function mutations of rl significantly reduced the suppressor activity of Dsor1Su1 (Table 1), demonstrating that Rl acts downstream of Dsor1 in the Tor pathway.
A temperature-sensitive gain-of-function allele of tor, torRL3 (Klingleret al. 1988), was significantly enhanced by rlSu23 and rlSu14. As shown in Table 2, considerable numbers of embryos produced by females heterozygous for torRL3 and also for either rlSu23 or rlSu14 failed to hatch at 28°. Most of the nonviable embryos had a reduced number of abdominal segments (data not shown).
The above results demonstrate the involvement of Rl in the Tor signaling pathway, and it is likely that the gain-of-function mutations of rl suppress the terminal defects of the terminal class mutant embryos similar to the gain-of-function mutations of Dsor1 (Tsudaet al. 1993; Limet al. 1997). Contrary to expectations, rlSu23 and rlSu14 did not suppress fs(1)ph at all: Neither the cuticular pattern nor the expression patterns of tll and hkb in the embryos lacking maternal fs(1)ph were affected by rlSu23 or rlSu14 (Table 1; Figure 4, E, J, and O).
Functions of rolled and Dsor1 in cell proliferation: Loss-of-function rl mutants die as third instar larvae that lack imaginal discs (Hilliker 1976), suggesting the involvement of rl in the proliferation of imaginal disc cells. Animals hemizygous for loss-of-function mutations of D-raf showed a similar phenotype (Perrimonet al. 1985; Nishidaet al. 1988; Tsudaet al. 1993; Hataet al. 1994). Clonal analysis demonstrated that the rate of proliferation was reduced ∼40% in clones homozygous for null D-raf1 (Tsudaet al. 1993; Hataet al. 1994; Table 3). A clonal analysis in the present study demonstrated that proliferation was much more severely affected in null Dsor1Gp158 than in null D-raf1 (Table 3). The rate of proliferation in Dsor1Gp158 was greatly reduced, and only one or two doubling events took place in most of the Dsor1Gp158 clones, while more than six doublings occurred in their sibling clones. Proliferation rates in the clones homozygous for the Dsor1 hypomorphs, Dsor1r1 and Dsor1r2, were affected in a similar manner to those observed in D-raf1 (Tsudaet al. 1993; Table 3).
To elucidate the function of rl in cell proliferation, we tested whether rlSu23 suppresses the proliferation defects in the D-raf and Dsor1 mutant clones. As shown in Table 3, rlSu23 suppressed the proliferation defects in the D-raf and Dsor1 mutant clones, although it did not restore the viability of the flies hemizygous for D-raf1, Dsor1r1, or Dsor1r2. This indicates that Rl acts downstream of Dsor1 in the signaling pathway regulating the imaginal cell proliferation. It should also be noted that the proliferation defects in null Dsor1Gp158 were significantly suppressed by rlSu23, suggesting considerable basal level activity by rlSu23.
Function of rolled in the Torso signaling pathway: Genetic and molecular characterization of dominant suppressors of D-raf on the second chromosome identified two gain-of-function alleles of rl with different dominant activities: rlSu14 (weak) and rlSu23 (strong). Surprisingly, the molecular lesion associated with rlSu23 was found to be identical to that of rlSem (Brunneret al. 1994). In fact, the dominant activity of rlSu23 on the eye and wing morphology was similar to that observed in rlSem. It has been reported that rlSem is invariably dominant female sterile with a dominant tor-like phenotype (Brunneret al. 1994), whereas both rlSu23 and rlSem showed only a weak dominant female sterility in this study. This discrepancy would be due to a difference in the genetic background: either the presence of a mutation(s) that enhances Tor signaling in the original rlSem stock or the occurrence of a suppressor mutation(s) in the current stocks of both rlSem and rlSu23. So far, outcrossings of the current stocks have shown no evidence of suppressor mutations.
The cell fates at the anterior and posterior termini of the early embryo are determined by the Tor signaling pathway (Nüsslein-Volhardet al. 1987; Luet al. 1993b). Unfortunately, the significance of rl in the pathway could not be directly assessed with germline clones because of its proximity to the centromeric heterochromatin (Berghella and Dimitri 1996). However, a half-reduction in the gene dosage of rl significantly reduced the suppressor activity of the dominant Dsor1 mutation on the posterior defect in the embryos lacking terminal class gene maternal activities (Table 1). This indicates that Rl is required in Tor signaling and that it acts downstream of Dsor1 (Figure 5). A detailed analysis of the gain-of-function mutations of rl demonstrated that they significantly strengthened the signals from the Tor receptor tyrosine kinase. The dominant activity is rather moderate and explains the observed weak dominant sterility (Figure 4 and Table 2). It also should be noted that the dominant rl mutations exhibited no suppressor activity on the terminal class mutations, in contrast to the dominant Dsor1 mutations (Figure 4; Table 1). The significance of this observation is discussed in a later section.
MAPK cascade and cell proliferation: The clonal analysis demonstrated that D-raf and Dsor1 encode the essential components of the signaling pathway regulating proliferation of imaginal disc cells (Tsudaet al. 1993). The proliferation defects of loss-of-function mutations of D-raf and Dsor1 were significantly suppressed by rlSu23, indicating that Rl acts downstream of Dsor1. It should also be noted that the proliferation defect in the null Dsor1Gp158 clones is much more severe than in the null D-raf1 clones (Table 3). This may suggest that D-raf mediates only a portion of the signals for proliferation to Dsor1 and that Dsor1 integrates growth-stimulating signals from other unknown factor(s) as well (Figure 5C).
Signal branching and integration in the MAPK cascade: As described above, the dominant rl mutations exhibited no suppressor activity on the terminal class mutations. This could be explained if the rl gain-of-function mutations were devoid of constitutive activity, and if the expression of their dominant activity were strictly dependent on upstream signals. However, the significant suppressor activity of rlSu23 in the proliferation defects in the null Dsor1Gp158 clones and in the R7 cell fate decision in the null sev mutants may run counter to this assumption.
It has been reported that the increased signal sensitivity of the mammalian ERK2D319N protein that has a mutation analogous to RlSem is due to a decreased sensitivity to dual-specificity MAPK phosphatases such as PAC1, CL100/MKP-1, MKP-2, and MKP-3 rather than to an increased kinase activity (Bottet al. 1994; Chuet al. 1996; Campset al. 1998). On the other hand, an in vitro kinase assay of the recombinant RlSem mutant protein produced in bacteria demonstrated significant activity for the phosphorylation of Yan, a native substrate of Rl, in the absence of activating MAPKK, while the normal recombinant Rl did not (Oellers and Hafen 1996). In the presence of activated mammalian MAPKK, RlSem exhibited a higher kinase activity than Rl+. The latter observations suggest an increased basal level activity of RlSem in addition to an increased sensitivity to the activator. The constitutive activity observed in vitro is consistent with the significant suppressor activity of rlSu23 in the proliferation defect in the null Dsor1Gp158 clones (Table 3). Taking these observations into account, it is most likely that the dominant activity of the RlSem and RlSu23 mutant proteins is due to both an increased basal level activity and a decreased sensitivity to inactivating phosphatases.
On the basis of the above considerations, we propose that the activation of Rl is necessary but not sufficient for Tor signaling, and that Dsor1 may provide yet another branching point in the Tor signaling pathway. One possible model would be that Dsor1 activates another unknown factor in addition to Rl in the Tor pathway, and that both are required for the transcriptional activation of tll and hkb (Figure 5A). It would also be possible that an inactivation of a factor that antagonizes the Rl function by Dsor1 would be required for the activation of the pathway (Figure 5B). Defects of varying degrees were seen in mitoses in the syncytial blastoderm embryos devoid of the maternal Dsor1 activity (L. Tsuda, H.-Y. Ha and Y. Nishida, unpublished observations), suggesting that Dsor1 participates in the regulation of mitosis and is activated throughout the embryo during cleavage divisions. Bifurcation of the Tor signals downstream of Dsor1 may constitute a mechanism for preventing Dsor1 from activating the target genes in regions other than the terminal regions of the embryo. As discussed above, integration of signals for imaginal cell proliferation would then take place at some other point in the MAPK cascade (Figure 5C). The differential branching and integration of signals may contribute to the functional diversification of the ubiquitous MAPK cascade.
We are grateful to G. Struhl, D. Brunner, E. Hafen, D. Yamamoto, and the Bloomington Stock Center for fly stocks. We are also greatly indebted to S. Tokumasu, T. Tsuboi, and K. Dohmoto for technical assistance. This work was supported by grants from the Mitsubishi Foundation, the Ministry of Education, Science, Sport, and Culture of Japan, and the Japan Science and Technology Corporation.
Communicating editor: N. Takahata
- Received April 3, 1999.
- Accepted June 8, 1999.
- Copyright © 1999 by the Genetics Society of America