ABSTRACT
The Drosophila segmentation gene paired, whose product is homologous to the Drosophila Gooseberry and mammalian Pax3 proteins, has three general functions: proper development of the larval cuticle, survival to adulthood and male fertility. Both DNA-binding domains, the conserved N-terminal paired-domain and prd-type homeodomain, are required within the same molecule for all general paired functions, whereas a conserved His-Pro repeat located near its C terminus is a transactivation domain potentiating these functions. The C-terminal moiety of Paired includes two additional functional motifs: one, also present in Gooseberry and Pax3, is required for segmentation and cuticle development; the other, retained only in Gooseberry, is necessary for survival. The male fertility function, which cannot be replaced by Gooseberry and Pax3, is specified by the conserved N-terminal rather than the divergent C-terminal moiety of Paired. We conclude that the functional diversification of paired, gooseberry and Pax3, primarily determined by variations in their enhancers, is modified by adaptations of their coding regions as a necessary consequence of their newly acquired spatiotemporal expression.
INTRODUCTION
Position along the anteroposterior axis of the developing Drosophila embryo is initially defined by the sequential activities of four classes of segmentation genes – maternal coordinate genes and zygotic gap, pair-rule and segment-polarity genes – and the homeotic genes, which form a hierarchical network (Nüsslein-Volhard and Wieschaus, 1980; Baumgartner and Noll, 1990; Small and Levine, 1991; Peifer and Bejsovec, 1992; St Johnston and Nüsslein-Volhard, 1992). Genes within this hierarchy are expressed in progressively refined domains and thus define the position along this axis with increasing precision.
The paired (prd) gene, which belongs to the pair-rule class of segmentation genes (Nüsslein-Volhard and Wieschaus, 1980), encodes a transcription factor containing a paired-domain (PD) and an extended prd-type homeodomain (HD) in its N-terminal half (Bopp et al., 1986) and a His-Pro (PRD) repeat near its C-terminal end (Frigerio et al., 1986). Prd protein is initially expressed in a broad anterior stripe at the end of the thirteenth nuclear division of syncytial blastoderm (Gutjahr et al., 1993a). By mid-cellularization, Prd appears in an anterior dorsal patch and in a characteristic pair-rule pattern of seven stripes, which by cellular blastoderm are converted into 14 stripes spanning each parasegment boundary. During germ band extension, Prd expression decreases in the epidermal stripes but later accumulates in a few specific cells of the central nervous system (CNS) and certain head regions (Gutjahr et al., 1993a). Together with the other pair-rule genes, prd specifies position along each double-segmental repeat and activates the segment-polarity genes, including gooseberry (gsb), wingless (wg) and engrailed (en), which are expressed at a single-segment periodicity (DiNardo and O’Farrell, 1987; Ingham et al., 1988; Bopp et al., 1989). In prd mutant embryos, every other stripe of Gsb, Wg and En protein is abolished, which results in the loss of the posterior part of even-numbered parasegments and of the adjacent anterior part of odd-numbered parasegments (Nüsslein-Volhard and Wieschaus, 1980; Bopp et al., 1989; Ingham and Martinez Arias, 1992).
As a transcription factor, Prd is of particular interest because it possesses in its N-terminal portion two DNA binding domains, a PD and a HD (Bopp et al., 1986; Treisman et al., 1991), while most transcription factors contain only one DNA binding domain. The coexistence of a PD and HD was also observed in several other members of the Pax gene family (Noll, 1993). Although in vitro experiments suggest that these two domains can function either independently or cooperatively when present in the same molecule (Treisman et al., 1991; Underhill et al., 1995; Jun and Desplan, 1996), the situation in vivo remains unknown. To understand the biological significance of the coexistence of the PD and HD in Prd, their role had to be examined in vivo.
We have previously shown that Prd is required in vivo not only for the expression of segment-polarity genes and normal development of the larval cuticle but also for the survival of the embryo to adulthood and for male fertility (Xue and Noll, 1996; Xue and Noll, 2000). Two Prd homologs, the Drosophila Gsb and murine Pax3 proteins, which share with Prd a highly conserved N-terminal moiety including the PD and HD, but have divergent C-terminal portions, are able to perform some of these functions of Prd when placed under the control of the entire prd cis-regulatory region (Xue and Noll, 1996). It follows that the acquisition of new cis-regulatory elements rather than changes in the coding region is the major evolutionary drive for the functional diversification among these three genes. However, both prd-Gsb and prd-Pax3 perform the cuticle function of Prd at low efficiency, and only prd-Gsb is able to rescue the prd mutant embryos to adulthood when it is present in two copies (Xue and Noll, 1996). These results indicate that the coding region also plays an important role in further modification of protein functions.
Here we show which protein domains of Prd are required for each of its in vivo functions. By constructing a series of prd transgenes expressing wild-type or mutated Prd proteins under the control of the entire prd cis-regulatory region and introducing them into prd mutants, we determined which of the Prd functions can be rescued by the respective transgenes. Our results demonstrate that both the PD and HD of Prd have to be present in cis for the activation of segment-polarity genes, wild-type cuticle formation and viability. In addition to the PRD repeat, which constitutes an important activation domain facilitating all functions of Prd, its C terminus contains at least two essential functional motifs. One motif, required for its function in larval cuticle development, is present in the C termini of both Gsb and Pax3, whereas another motif, needed for its role in viability, is present only in the Gsb C terminus. Finally, the determinant for the male fertility function, which cannot be replaced by the prd-Gsb and prd-Pax3 transgenes, resides in the conserved N-terminal rather than the divergent C-terminal moiety of Prd. This observation strongly challenges the classic view using percentage amino acid identity as a measure of functional equivalence between homologous proteins.
MATERIALS AND METHODS
Construction of rescue and expression plasmids
All transgenes whose coding portions are illustrated in Fig. 1 were generated in two steps. First, wild-type cDNAs or their mutated versions were subcloned into pKSpL5, a derivative of Bluescript pKS+ (Xue and Noll, 1996), or into the appropriate pAR vector (Studier and Moffat, 1986). Subsequently, these sequences were recovered as XbaI fragments from pKSpL5 or XbaI-NheI fragments from pAR subclones and inserted in the correct orientation into the SpeI site of the prd-0 P element construct (Xue and Noll, 1996).
To produce the pKSpL5-Gsb, pKSpL5-Prd and pKSpL5-Pax3 subclones, the gsb-cDNA BSH9c2 (Baumgartner et al., 1987), prd-cDNA c7340.6 (Frigerio et al., 1986) and Pax3-cDNA (Goulding et al., 1991), respectively, were cloned into the unique EcoRI site of the polylinker of pKSpL5 in the required orientation. To obtain pKSpL5-GsbP17L, the 1.4 kb EagI-HindIII fragment in pKSpL5-Gsb was replaced by the corresponding PCR product, generated with the primers gsbP17L (5′-TTC ATC AAC GGC CGT CTG TTG-3′) and T3 (5′-ATT AAC CCT CAC TAA AG-3′). To generate pKSpL5-int-Gsb, the prd intron was amplified by PCR from prd-SN20 with the following primers (mismatches were introduced to generate the underlined SpeI sites): prdint1 (5′-GAT ATT CTA CTA GTC AAG GTG AG-3′) and prdint2 (5′-GCC GCT GTA CTA GTC TGG AATGA-3′). Subsequently, the PCR product was inserted into the unique SpeI site of the polylinker of pKSpL5-Gsb in the appropriate orientation. The pKSpL5-PrdΔPRD construct was produced by inserting the blunt-ended 2.2 kb HindIII-EcoRI fragment from PrdΔPR (Cai et al., 1994) between the two SmaI sites in the pKSpL5 polylinker. To obtain pKSpL5-PrdN+GsbC, the 850 bp EcoRI-PvuII fragment of c7340.1 prd-cDNA (Frigerio et al., 1986) and the 750 bp FspI-EcoRI fragment of pKSpL5-Gsb were ligated into the EcoRI site of pKSpL5. To generate pKSpL5-PrdN, the 2.4 kb EcoRI-SpeI fragment in pKSpL5-Prd was replaced with the 1.0 kb EcoRI-AvrII fragment obtained by PCR from pKSpL5-Prd with the use of the primers T7 (5′-AAT ACG ACT CAC TAT AG-3′) and prd-4 (5′-GCC TGA GAC CTA GGT GTG CTG-3′).
pAR-GsbΔP, pAR-GsbΔH, pAR-GsbC and pAR-GsbN were produced by PCR mutagenesis and subcloned into the pAR3040 vector (X. Li, L. X. and M. N., unpublished). To generate pAR-Gsb+PRD, the 650 bp SacI-NheI fragment from pAR-gsb.fl (Gutjahr et al., 1993b) was replaced with the corresponding 950 bp fragment from a prd-cDNA subcloned in pAR3038 (Gutjahr et al., 1993a), pAR-prd. To construct pAR-GsbN+PoxnC, the NdeI/SpeI-digested GsbN PCR product (amplified from pAR-Gsb with the primers T7 and gsbres8 [5′-CCT GCT GGG TGA CTA GTT GCT TGC GCA-3′]) was ligated with the SpeI/BclI-digested PoxnC PCR product (amplified from the poxn-cDNA P4c6 [Dambly-Chaudière et al., 1992] with the primers poxnres3 [5′-TCA AAA CTT GAT CAG TGG CGA GA-3′] and poxnres4 [5′-GCG CAA CAG CGG ACT AGT GAC CGA TGA GAT-3′]) between the NdeI and BamHI sites of the pAR3040 polylinker. To generate pAR-GsbN+Pax3C, the NdeI/SpeI-digested GsbN PCR product and the SpeI/BamHI-digested Pax3C PCR fragment (amplified from pKSpL5-Pax3 with the primers pax3res2 [5′-CCA GGA GGA TCC ACC CCC TAG AAC GT-3′] and pax3res3 [5′-TGG AGG AAA CTA GTT GGA GCC AA-3′]) were ligated between the NdeI and BamHI sites of pAR3040, whereas pAR-GsbN+PrdC was obtained by ligating the NdeI/SpeI-digested GsbN fragment and the NheI/BclI-digested PrdC PCR product (amplified from pKSpL5-Prd with the primers prdC-1 [5′-CGC AAG CAG CTA GCC TCG GTC TC-3′] and prdC-2 [5′-GTA GGT GGT TGA TCA GTG TCT CT-3′]) between the NdeI and BamHI sites of pAR3040. Finally, the pAR-Pax3N+GsbC was generated by inserting the NdeI/BamHI-digested Pax3N PCR product (amplified from pKSpL5-Pax3 with the primers pax3res4 [5′-GCT GCC CCC CAT ATG ACC ACG CT-3′] and pax3res5 [5′-AGT TGA TTG GAT CCA GCT TGT-3′]) and the BamHI-NheI fragment of pAR-GsbC between the NdeI and NheI sites of the pAR3040 vector.
Generation of transgenic flies
Rescue constructs were injected together with pUChspΔ2-3 P-element helper plasmid (D. Rio, personal communication) into ry506 embryos, and ry+ transformants were selected (Rubin and Spradling, 1982).
Transgenic prd− embryos carrying one or two copies of the specified transgenes were obtained as follows. Two types of stocks were established for all transgenes except prd-GsbN, prd-GsbN+PoxnC and prd-PrdN, which exhibit a dominant-negative effect on Prd functions, namely Df(2L)Prl/SM1; P/P and prd2.45/SM1; P/P (P stands for the P elements that contain the transgenes). To rescue the cuticle functions of Prd with one or two copies of the transgenes, prd mutant embryos were collected from crosses between prd2.45/SM1; P/P and prd2.45/SM1 flies or from crosses of prd2.45/SM1; P/P flies inter se. To rescue the viability and male fertility functions of Prd with one or two copies of the transgenes, prd2.45/SM1; P/P flies were crossed with Df(2L)Prl/SM1 or Df(2L)Prl/SM1; P/P flies. When the transgene failed to rescue the cuticle functions, its ability to provide the viability functions was assayed by supplying the cuticle functions with two copies of the prd-Pax3 transgene. To rescue the male fertility function of Prd with one or two copies of the prd-PrdN (or prd-GsbN) transgene, male flies of the genotype Df(2L)Prl prd-GsbN+PrdC/prd2.45prd-GsbN+PrdC; prd-PrdN (or prd-GsbN)/TM3, Sb or Df(2L)Prl prd-GsbN+PrdC/prd2.45prd-GsbN+PrdC; prd-PrdN (or prd-GsbN)/prd-PrdN (or prd-GsbN) were obtained from crosses between Df(2L)Prl prd-GsbN+PrdC/SM1; prd-PrdN (or prd-GsbN)/TM3, Sb and prd2.45prd-GsbN+PrdC/SM1; prd-PrdN (or prd-GsbN)/TM3 Sb flies and subsequently tested for fertility. To rescue the male fertility function of Prd with one or two copies of the prd-GsbN+Pax3C transgene, one copy of prd-Gsb was supplied to rescue the viability functions by crossing Df(2L)Prl prd-GsbN+Pax3C/SM1; prd-Gsb flies with prd2.45/SM1 or prd2.45prd-GsbN+Pax3C/SM1 flies. The prd-GsbN+PoxnC transgene could not be tested for the viability and male fertility functions of Prd because of its strong dominant-negative effect, which results in lethality in the presence of one copy of the wild-type prd gene.
Expression and purification of proteins and band-shift assays
Prd, Gsb, GsbΔP, GsbΔH, GsbC, GsbN, GsbN+PoxnC, GsbN+Pax3C and GsbN+PrdC proteins were expressed in E. coli BL21(DE3) cells, transformed with the corresponding pAR subclones and purified to about 50% purity as described previously (Gutjahr et al., 1993a). Band-shift assays were performed as described (Xue and Noll, 1996).
Immunostaining of embryos and preparation of cuticles
Immunostaining of embryos with the anti-Gsb antiserum was carried out as described (Gutjahr et al., 1993b). Double-staining of embryos for β-gal and Gsb, Wg or En protein was performed according to Lawrence et al. (Lawrence et al., 1987). Cuticles were prepared essentially as described (Wieschaus and Nüsslein-Volhard, 1986).
RESULTS
Both PD and HD are strictly required in cis for the functions of Prd
The independent contributions of the PD and HD to the in vivo functions of the Prd protein were investigated with the aid of flies carrying the prd-Gsb transgene, which expresses the Gsb protein under the control of the entire prd cis-regulatory region and performs all Prd functions required for survival of the embryos to adults (Xue and Noll, 1996). Three rescue constructs, prd-GsbΔP, prd-GsbΔH and prd-GsbC, were derived from prd-Gsb by deleting the PD, the extended prd-type HD, or both (Fig. 1). More than 10 independent transgenic lines were obtained for each construct. Each transgene was subsequently introduced into prd mutants and tested for its ability to rescue the following functions of Prd: (i) activation of segment-polarity genes downstream of prd, such as gsb, wg and en, (ii) development of wild-type larval cuticle, and (iii) survival to adulthood after rescue of the cuticular phenotype by prd-Pax3. The first two tests assay for what we will henceforth call the cuticle function of Prd, the third for its viability function (Xue and Noll, 1996). The prd embryos used in these and all subsequent experiments were either homozygous for the prd2.45 allele or hemizygous when this allele was combined with the deficiency Df(2L)Prl. The prd2.45 allele does not produce any functional Prd protein because it carries a 1.1 kb insertion (Kilchherr et al., 1986) after amino acid 45 of the PD (Frigerio et al., 1986). The tests showed that none of the three transgenes is able to perform any function of Prd, even when two copies are present (Figs 1-4).
In discussing these results the following findings are relevant. First, all three transgenes express truncated Gsb proteins at a level comparable to that expressed by the prd-Gsb transgene and detected by anti-Gsb specific for the C-terminal moiety (Gutjahr et al., 1993b). Second, homozygous prd-GsbΔP exerts a weak dominant-negative effect on the cuticle function of wild-type Prd, i.e., a few embryos show a weak pair-rule phenotype while most embryos are wild-type and survive to adulthood. Third, GsbΔH and GsbC proteins are able to perform the Gsb function in cuticle development when expressed under the control of gsb cis-regulatory elements (X. L., L. X. and M. N., unpublished). Consequently, the products of the transgenes are made and are partially functional, but they cannot rescue any of the Prd functions. Evidently, both the PD and HD are required for the in vivo functions of Prd.
To minimize the possible disruption of the overall protein structure that might result from the deletion of entire domains, we constructed prd-GsbP17L, which contained a PD in which Pro at position 17 was replaced by Leu (Fig. 1), identical to the amino acid substitution resulting from point mutations in human PAX3 and PAX5. The first human mutation produces Waardenburg’s syndrome I (Baldwin et al., 1992; Tassabehji et al., 1992), the second results in a complete loss of its DNA-binding activity in vitro (Czerny et al., 1993). The prd-GsbP17L, like prd-GsbΔP, fails to rescue any Prd functions when similarly tested (Figs 1 and 3D,G) and hence suggests that DNA binding of the PD is essential for the rescue of Prd functions.
When the PD, HD or both are deleted in Gsb, it can no longer bind to GEE1 (Xue and Noll, 1996), a Prd target site located in the gsb early enhancer GEE (Li et al., 1993), as evident from band-shift assays (Fig. 5, lanes 8-10). This is in agreement with the in vivo results and implies that both domains recognize binding sites in GEE1 with which they probably interact cooperatively.
To test whether prd-GsbΔP and prd-GsbΔH can complement each other to rescue some Prd functions, we introduced both transgenes into the same prd mutants in four different combinations: one copy or two copies of each transgene or two copies of one combined with one copy of the other. In all four combinations, no rescue is observed of segment-polarity gene activation (Fig. 2P-R), cuticular phenotype (Fig. 3H) and viability (when the cuticle function was provided by prd-Pax3; not shown). It follows that prd-GsbΔP and prd-GsbΔH cannot complement for any Prd functions. Consistent with this result, GsbΔP and GsbΔH proteins when present together failed to interact with GEE1 in a band-shift assay (Fig. 5, lane 11). We conclude that both the PD and HD are strictly required in cis for the in vivo functions of Prd, most probably because of their cooperative DNA-binding activities.
The prd intron is dispensable for the normal functions of Prd
From the original prd rescue construct prd-SN20, capable of performing all Prd functions (Gutjahr et al., 1994), the prd-Gsb transgene was derived by replacing a genomic fragment including the end of the leader, most of the coding region and the 356 bp intron of prd with a gsb-cDNA (Xue and Noll, 1996). As indicated by our previous results, activation of Prd target genes and the cuticular phenotype of prd− embryos could be fully rescued with two and partially with one copy of the prd-Gsb transgene. In addition, the prd-Gsb transgene was able to rescue some prd mutants to adulthood when present in two copies. These results demonstrated that the prd-Gsb transgene, albeit at a lower efficiency than prd-SN20, is able to perform all the Prd functions that are required by the animal to survive to adulthood. Further studies showed that all male flies rescued by the prd-Gsb transgene are sterile, presumably because the severely reduced accessory glands have lost their functions (Xue and Noll, 2000). Therefore, a new function of Prd, henceforth called the male fertility function, was uncovered by the presence of the prd-Gsb transgene in prd mutants. This conclusion was confirmed by the observation that a prd rescue construct, differing from prd-SN20 by the lack of 5 kb of the downstream regulatory sequences, is unable to confer fertility on prd mutant males rescued to viable adults since these lack accessory glands (Bertuccioli et al., 1996; Xue and Noll, 2000).
These results left open the question of whether the reduced ability of prd-Gsb to replace the prd functions described above are attributable to the protein coding region or to the presence of enhancers in the missing prd intron. To distinguish between these possibilities, we constructed two transgenes (Fig. 1). The prd-Prd transgene was derived from prd-SN20 by deleting the intron. In the other transgene, prd-int-Gsb, the prd intron was inserted into prd-Gsb between the prd leader and the gsb-cDNA sequences. We found that one copy of the prd-Prd transgene is sufficient to fully rescue all Prd functions in a prd−background, namely cuticle development (Fig. 6A), viability (Fig. 4) and male fertility (Fig. 1). On the other hand, the prd-int-Gsb transgene is functionally indistinguishable from prd-Gsb as in its presence the cuticular phenotype of prd mutants is rescued partially by one copy (Fig. 6B) and completely by two copies (Fig. 1). In addition, two copies of the prd-int-Gsb transgene can also rescue about 10% of prd mutants to adulthood (Fig. 4), an efficiency comparable to that of prd-Gsb (Fig. 4; Xue and Noll, 1996). By contrast, all rescued males are sterile (Fig. 1). From these results, we conclude that the prd intron is dispensable for the normal functions of prd.
The PRD repeat constitutes an important transactivation domain
In addition to the PD and HD, the prd gene encodes a third domain, the 21-amino acid His-Pro (or PRD) repeat, located near its C-terminal end (Frigerio et al., 1986). To investigate the in vivo function of the PRD repeat, we constructed two transgenes: prd-PrdΔPRD and prd-Gsb+PRD (Fig. 1). In prd-PrdΔPRD, the PRD repeat was removed by PCR mutagenesis, while in prd-Gsb+PRD the last 12 amino acids of Gsb were replaced by the 118-amino acid C-terminal end of Prd, which includes the PRD repeat. Although one copy of the prd-PrdΔPRD transgene can fully rescue the cuticular phenotype of prd mutants (Fig. 6C), its rescue efficiencies for viability (Fig. 4) and male fertility (Fig. 1) are dramatically reduced. Thus, one copy of prd-PrdΔPRD rescues less than 20% of prd mutants to viable adults, an efficiency that is 4-to 6-fold lower than that seen with prd-Prd (Fig. 4), and the fertility of the rescued males is strongly reduced (Fig. 1). However, two copies of prd-PrdΔPRD completely rescue both the embryonic lethality (Fig. 4) and male sterility of prd mutants (Fig. 1).
Moreover, prd-Gsb+PRD displays considerably enhanced rescue efficiencies of the cuticle phenotype and embryonic viability as compared to those observed with prd-Gsb. One copy of prd-Gsb+PRD completely rescues the cuticular phenotype of prd mutants (Fig. 6D). It can also rescue the embryonic lethality, partially with one copy and fully with two copies, an efficiency that is about 9-fold higher than that of prd-Gsb (Fig. 4). However, neither one nor two copies of prd-Gsb+PRD are able to rescue the male sterility of prd mutants (Fig. 1).
Evidently, the PRD repeat, though not absolutely required, potentiates all functions of Prd, and thus functions as an important transcriptional activation domain.
The C-terminal portion of Prd includes at least three functional motifs
Our previous work showed that two evolutionary alleles of prd, prd-Gsb and prd-Pax3, which share the same cis-regulatory region but not identical coding sequences with prd, have conserved most of the functions of Prd (Xue and Noll, 1996). However, neither Gsb (Xue and Noll, 2000) nor Pax3 are able to replace the male fertility function of Prd, as all prd− males rescued by two copies of prd-Gsb or combinations of prd-Gsb with prd-Pax3 are sterile (Fig. 1). These results demonstrated that the Prd, Gsb and Pax3 proteins are functionally nearly equivalent. Although no significant similarity has been found among their C-terminal sequences, it is probable that some functional motifs in their 3-D structures have been conserved during evolution (Xue and Noll, 1996). Thus, the C terminus of Prd may contain, in addition to the PRD repeat, three motifs (‘motif’ defined as a single feature of a domain that permits a specific interaction of the domain with another protein or nucleic acid, and thus may consist in extreme cases of a single specific amino acid within a domain, such as position 50 of the homeodomain) or domains that are necessary for the cuticle, viability and male fertility functions of Prd. While both prd-Gsb and prd-Pax3 can support normal cuticle development, only prd-Gsb is able to rescue the lethality of prd mutants (Xue and Noll, 1996). Hence, we assume that the motif required for the cuticle function is conserved among the C termini of Prd, Gsb and Pax3, whereas the motif essential for viability is expected to be present only in the C termini of Prd and Gsb. In addition, as both Gsb and Pax3 are unable to replace Prd in promoting male fertility, the motif performing this function is apparently missing in both Gsb and Pax3.
To test this hypothesis, we constructed a series of transgenes, prd-GsbN, prd-GsbN+PoxnC, prd-GsbN+Pax3C and prd-GsbN+PrdC, all derived from prd-Gsb by either deleting its C-terminal moiety or replacing it with that of Pox neuro (Poxn) (Dambly-Chaudière et al., 1992), Pax3 or Prd (Fig. 1). All these proteins bind, in band-shift assays, to the Prd target site GEE1 (Xue and Noll, 1996) with affinities similar to those of Prd and Gsb (Fig. 5, lanes 2-7), presumably by their two DNA-binding domains of GsbN. If the functional divergence observed for prd-Pax3, prd-Gsb and prd-Prd resides in the C-terminal portions of their coding regions, we expect (i) that prd-GsbN and prd-GsbN+PoxnC lack the hypothetical motifs required for Prd-dependent cuticle formation, viability and male fertility, and thus are unable to perform any functions of Prd; (ii) that prd-GsbN+Pax3C, which is assumed to contain the motif for the cuticle function, can rescue the cuticle phenotype, but not the lethality of prd mutants; and (iii) that prd-GsbN+PrdC, which includes all functional motifs present in the C terminus of Prd, is able to execute all Prd functions. Most of these expectations were borne out by the following findings. First, either one or two copies of prd-GsbN or prd-GsbN+PoxnC are unable to provide the cuticle function (Fig. 1). Moreover, both prd-GsbN and prd-GsbN+PoxnC exert a dominant-negative effect on the cuticle function of Prd since one copy of prd-GsbN+PoxnC or two copies of prd-GsbN are lethal and produce a pair-rule phenotype when only one wild-type allele of prd is present (not shown). This effect might result from formation of inactive heterodimers of these proteins with Prd through their HDs or from competition for Prd DNA target sites of GsbN or GsbN+PoxnC, as these proteins possess similar DNA-binding, yet not the same transcriptional activation capabilities as Prd. Second, prd-GsbN+Pax3C, like prd-Pax3, is able to rescue, partially with one copy (Fig. 6E) and completely with two copies (Fig. 1), the cuticular phenotype of prd mutants, yet fails to rescue lethality (Fig. 4) and male sterility (Fig. 1) of these mutants. Finally, as expected, one copy of prd-GsbN+PrdC can fully rescue all Prd functions that are required for embryonic survival to adulthood (Figs 4, 6F). However, surprisingly prd-GsbN+PrdC fails to rescue the male sterility of prd mutants (Fig. 1), which implies that at least part of the motif required to promote male fertility resides in the highly conserved N-terminal moiety of Prd.
We conclude that, in addition to a transactivation domain, the C-terminal portion of Prd includes sequences important for its cuticle and viability functions. Furthermore, the motif required for cuticle formation is retained in the C termini of both Gsb and Pax3, the viability function only in that of Gsb.
Mapping of viability and cuticular functions to the C-terminal moiety of Prd
An approximate location in Prd of the motif required for its cuticle function may be determined from the prdIIN mutant (Tearle and Nüsslein-Volhard, 1987). Since this prd allele produces a weak cuticular phenotype and truncates the 126 C-terminal amino acids of Prd (Bertuccioli et al., 1996), part of the motif required for the cuticle function is located in this C-terminal portion of Prd (Fig. 7). In addition, another truncated Prd protein, PrdΔPRT, lacking the last 74 amino acids including the PRD repeat, is still able to promote normal cuticle formation (Bertuccioli et al., 1996). Therefore, the motif required for cuticle function must be located around residue 487 and before residue 540 of Prd (Fig. 7).
To map the motif necessary for the viability function, we introduced either one or two copies of the prd-Pax3 transgene into Df(2L)Prl/prdIIN mutants. While one copy of the prd-Pax3 transgene is unable to rescue any mutants to adulthood, two copies of prd-Pax3 do rescue some of these mutants to viable adults (not shown), yet cannot rescue Df(2L)Prl/prd2.45 embryos, which lack any functional Prd protein (Xue and Noll, 1996). This suggests that at least part of the sequences needed for viability must reside between the beginning of the C-terminal portion of Prd at residue 273 and the end of PrdIIN at residue 487 (Fig. 7).
The male fertility function is determined by the N-terminal portion of Prd
Although the N-terminal portions of Prd, Gsb and Pax3 are highly conserved and exhibit the same DNA-binding ability in vitro (Xue and Noll, 1996), the failure of prd-GsbN+PrdC to rescue the male sterility of prd mutants implies that they are functionally different. To corroborate this conclusion, two transgenes were constructed, prd-Pax3N+GsbC and prd-PrdN+GsbC (Fig. 1), in which the N-terminal portion of Gsb in prd-Gsb is replaced by that of Pax3 or Prd. In prd mutants, prd-Pax3N+GsbC behaves like prd-Gsb. It is able to rescue the cuticular phenotype, partially with one copy (Fig. 6G) and fully with two copies (Fig. 1). In particular, two copies of prd-Pax3N+GsbC can rescue some prd mutants to viable adults with an efficiency comparable to that of prd-Gsb (Fig. 4), but fail to rescue the sterility of male flies (Fig. 1). These results indicate that the N-terminal portions of Gsb and Pax3 are equivalent with respect to the functions of Prd. By contrast, prd-PrdN+GsbC is not only able to rescue the cuticular phenotype (Figs 1, 6H) and lethality (Fig. 4) of prd mutants with similar efficiencies as prd-Gsb, but also the male fertility function of Prd (Fig. 1). It follows that the N-terminal portion of Prd includes the crucial determinant for its role in male fertility (Fig. 7), presumably through its recognition of specific DNA target sites. The fact that prd-PrdN+GsbC needs two copies to rescue male sterility is consistent with the finding from prd mutants, rescued by prd-PrdΔPRD, that the PRD repeat is required to enhance the efficiency of the protein in promoting male fertility (Fig. 1).
To investigate if the N-terminal portion of Prd suffices to rescue male sterility, the prd-PrdN transgene was constructed by deleting the C-terminal portion of prd-Prd (Fig. 1). The cuticle and viability functions of prd-PrdN were examined in prd mutants, its male fertility function in prd mutants rescued by two copies of prd-GsbN+PrdC. Not surprisingly, prd-PrdN is unable to rescue any Prd functions (Figs 1, 4). On the contrary, prd-PrdN exerts a dominant-negative effect on prd since two copies of prd-PrdN are lethal and produce a pair-rule phenotype in the presence of only one wild-type prd allele (not shown), presumably because PrdN inhibits Prd function by competing with Prd for the same DNA target sites or by forming inactive heterodimers with Prd.
In summary, these results show that (i) the function of Prd in male fertility is determined by its N-terminal portion, which can bind specifically to DNA targets not recognized by Gsb and Pax3; (ii) the N-terminal portion of Prd is by itself not sufficient to perform any Prd functions, presumably because of the lack of an activation domain; (iii) the N-terminal portions of Prd, Gsb and Pax3 are functionally divergent despite their high sequence conservation; and (iv) the C-terminal portions of Prd and Gsb are qualitatively similar in functions even though their primary sequences have widely diverged.
DISCUSSION
The Drosophila prd gene is of interest because it encodes a multidomain transcription factor with multiple functions and, in form of its paralogs gsb and gsbn and its orthologs Pax3 and Pax7 (Fitch, 2000), plays a key role in both invertebrate and vertebrate development (Noll, 1993). Here we discuss the results of an extensive analysis aimed at elucidating the contributions of the various domains of the Prd protein to its in vivo functions. The prd gene is particularly well suited for such studies because (i) its entire cis-regulatory region has been identified (Gutjahr et al., 1994); (ii) its in vivo functions have been characterized (Nüsslein-Volhard and Wieschaus, 1980; Xue and Noll, 1996, 2000); (iii) various prd mutants are available (Nüsslein-Volhard and Wieschaus, 1980; Nüsslein-Volhard et al., 1984; Kilchherr et al., 1986); and (iv) the complete rescue of its in vivo functions has been achieved (Gutjahr et al., 1994).
In contrast to previous work performed in vitro (Treisman et al., 1991; Underhill et al., 1995; Jun and Desplan, 1996) or in cell culture (Han et al., 1989; Cai et al., 1994), the present study is aimed at establishing all wild-type in vivo functions of the Prd protein, a task that requires expression under the original control elements rather than by the artificial means of a heat-shock promoter (Miskiewicz et al., 1996) or incomplete prd enhancer elements (Bertuccioli et al., 1996).
With the aid of two evolutionary alleles of prd, prd-Gsb and prd-Pax3, in which the gsb and Pax3 coding regions were placed under the control of the entire prd cis-regulatory region, it was shown previously that Prd activity is required in vivo during at least three distinct developmental stages to ensure proper segmentation of the larval cuticle, postembryonic viability and male fertility (Xue and Noll, 1996; Xue and Noll, 2000). Here, we constructed a series of prd transgenes which express various versions of the Prd protein, including truncations or chimeras of Prd, Gsb and Pax3 under the control of the complete prd cis-regulatory region. All transgenes were tested for their ability to rescue any of these Prd functions. Thus, this report is the first example of a complete functional analysis of the Prd protein under natural conditions.
Cooperativity between PD and HD
The presence of two DNA-binding domains, PD and HD, in Prd and some other members of the Pax gene family raises the question of whether the regulation of any of its target genes requires the binding of both or only one of its two DNA-binding domains. Both mechanisms are compatible with in vitro results (Treisman et al., 1991; Underhill et al., 1995; Jun and Desplan, 1996). Our in vivo studies show that both PD and HD are absolutely required for Prd function because deletion of either or both of these domains from the prd-Gsb transgene result in the complete loss of its ability to rescue the segment-polarity gene activation, cuticular phenotype and lethality of prd mutants. Moreover, since a point mutation in the PD, prd-GsbP17L, eliminates all Prd functions, the DNA-binding ability of the PD is necessary for the normal functions of Prd. An analogous mutation abolishes DNA binding of the human PAX5 protein (Czerny et al., 1993) and causes Waardenburg’s syndrom I when present in PAX3 (Baldwin et al., 1992). Our observation that prd-GsbΔP and prd-GsbΔH cannot complement for any function of Prd implies that the PD and HD must be present in the same Prd molecule, presumably because each Prd function requires the recognition of at least one composite DNA target site.
In agreement with our findings, Prd proteins unable to bind DNA as a result of single amino acid substitutions in either the PD or HD can no longer activate the ectopic expression of Prd-target genes when expressed ubiquitously under the control of the heat-shock promoter (Miskiewicz et al., 1996) nor will they perform any Prd in vivo function when expressed under the control of some of its own enhancers (Bertuccioli et al., 1996). In addition, a composite Prd target site has been identified in the even-skipped enhancer whose mutation in either the PD or HD binding portion dramatically reduces Prd binding activity both in vitro and in vivo (Fujioka et al., 1996). Our finding that the PD and HD cannot complement in trans for any function of Prd agrees with some observations obtained with mutant transgenes in vivo (Bertuccioli et al., 1996), but contradicts results obtained in vitro (Jun and Desplan, 1996), and in vivo when the two Prd mutant proteins are expressed under heat-shock control (Miskiewicz et al., 1996). Taken together, these results imply that the PD and HD of Prd may interact with their DNA targets cooperatively and that this cooperativity can occur in trans only if the proteins are produced at concentrations much higher than those occurring naturally.
The role of the C-terminal PRD repeat of Prd
The PRD repeat, which encodes a 20-30 amino acid His-Pro repeat, was discovered in an attempt to verify predictions of the gene network hypothesis in a search for protein-coding domains of prd (Frigerio et al., 1986; Noll, 1993). It was found in a number of Drosophila early developmental genes, including bicoid (bcd) and daughterless (da) (Frigerio et al., 1986; Berleth et al., 1988; Cronmiller et al., 1988), but its in vivo function remained unknown. Previous experiments in cell culture systems showed that the PRD repeat is part of a transactivation domain (Han et al., 1989; Cai et al., 1994) that is necessary to drive ectopic expression of Prd-target genes under the control of ubiquitously expressed Prd (Cai et al., 1994). Other studies, however, suggested that the PRD repeat is not essential for in vivo functions of Prd (Bertuccioli et al., 1996).
Our results demonstrate that the Prd protein whose PRD repeat has been deleted in prd-PrdΔPRD is still able to perform all in vivo functions of Prd, which implies that the PRD repeat is not absolutely required for Prd function. However, the fact that one copy of prd-PrdΔPRD exhibits significantly reduced efficiency in its ability to rescue the lethality and male sterility of prd mutants indicates that the PRD repeat greatly facilitates these Prd functions. This conclusion is corroborated and extended by the results obtained with prd-Gsb+PRD transgenes, which demonstrate that the PRD repeat enhances the viability as well as the cuticle function of Prd. Thus, the PRD repeat is an important transactivation domain that facilitates all functions of Prd.
Conservation of functional motifs in the divergent C-terminal moieties of Prd, Gsb and Pax3
Previous work has demonstrated that Prd, Gsb and Pax3 proteins are, at least partially, functionally equivalent (Li and Noll, 1994; Xue and Noll, 1996). When expressed under the control of the entire cis-regulatory region of prd, both Gsb and Pax3 can activate Prd-target genes necessary for the generation of wild-type cuticle, while Gsb is able to rescue prd mutants to adulthood. These results strongly suggested that the acquisition of cis-regulatory regions rather than the divergence of their coding regions is the primary evolutionary mechanism responsible for the functional diversification of prd, gsb and Pax3 genes. However, although Gsb and Pax3 can substitute for most Prd functions, they do so at considerably reduced efficiencies, which indicates that these proteins had to adapt their new functions for optimal performance by subsequent mutations producing the observed divergence of the Prd, Gsb and Pax3 proteins. Here, we have studied the result of this process of adaptation by examining the functional differences between these proteins when expressed as evolutionary alleles under the same cis-regulatory region.
Our results lead us to postulate that, in addition to the PRD repeat, two motifs or domains are present in the C-terminal portion of Prd on whose functions the formation of wild-type larval cuticle and survival to adulthood depend. Although no significant similarity has been found among the primary sequences of the C-terminal moieties of Prd, Gsb and Pax3, the motif required for implementing wild-type cuticle is shared by all three proteins. In contrast, the motif necessary for Prd’s viability function is retained only in Gsb, presumably as secondary or tertiary protein structure. It should be stressed that at least two independent functions of Prd are required for viability, one of which Pax3 is able to perform even better than Gsb (Xue and Noll, 1996). However, Pax3 is unable to substitute for one of the viablity functions of Prd and even exerts a dominant-negative effect on it (Xue and Noll, 1996). In agreement with our postulate, combining our results with those obtained with two weak prd alleles encoding truncated Prd proteins (Tearle and Nüsslein-Volhard, 1987; Bertuccioli et al., 1996), allowed us to map the motifs for the cuticle and viability functions within the C terminus of Prd.
The male fertility function of Prd
Although prd-Gsb, an evolutionary allele of prd, rescues prd mutants to viable adults, all males are sterile. Since wild-type males transgenic for two copies of prd-Gsb are fertile, we conclude that prd has a function required for male fertility. Moreover, as prd-Gsb includes the entire cis-regulatory region of prd (Xue and Noll, 1996), its failure to rescue male fertility must be caused by the inability of Gsb to replace this function of the Prd protein. Since Prd and Gsb share a highly conserved N-terminal portion consisting of two DNA-binding domains, the PD and HD, it seemed plausible to map this functional difference to their divergent C termini. Surprisingly, however, the protein-domain-swapping experiments indicate that the conserved N-terminal rather than the divergent C-terminal portion is the determinant for this particular function of Prd. Therefore, we suggest that at least one specific Prd target site, recognized by Prd but not Gsb, is involved in male fertility.
The male fertility function of Prd is controlled by a specific prd enhancer uncovered in prd mutants by a prd rescue construct that lacks 5 kb of the downstream regulatory region (Bertuccioli et al., 1996). Consistent with this interpretation, a prd transgene that expresses Prd merely under the control of this 5 kb regulatory region is able to confer fertility to prd-Gsb males mutant for prd (L. X. and M. N., unpublished). Males completely deficient for this fertility function of prd have no accessory glands, while accessory glands begin to form in prd mutant males rescued by prd-Gsb, but stop development at a severely reduced size (Bertuccioli et al., 1996; Xue and Noll, 2000). These findings are in agreement with our hypothesis that new functions evolve primarily through the acquisition of new enhancers during gene duplication (Li and Noll, 1994; Xue and Noll, 1996) and that the adaptation of the protein is secondary and a necessary consequence of its expression in the newly acquired context of this function.
Our results further imply that the C-terminal portions of Prd and Gsb, though divergent in their primary sequences, are still qualitatively the same. Hence, we questioned the validity of amino acid similarity as a general measure of functional equivalence in homologous proteins (Xue and Noll, 1996). Instead, Yockey (1992) proposed to replace this measure of functional equivalence by calculations of the mutual entropy between two protein sequences, a more precise statistical measure that takes into account the probability by which certain amino acids are replaced by others.
Acknowledgments
We are indebted to T. Kornberg for anti-En monoclonal antibodies, R. Nusse for anti-Wg antiserum, E. Hafen for the pUChspΔ2-3 P-element helper plasmid, P. Gruss for a Pax3-cDNA, M. Weir for the PrdΔPR plasmid and C. Nüsslein-Volhard for the prd mutant stocks. We thank Fritz Ochsenbein for expert art work, E. Frei for stimulating discussions and J. Alcedo, E. Frei and H. Noll for critical comments on the manuscript. This work was supported by Grant 31-40874.94 from the Swiss National Science Foundation and by the Kanton Zürich.