spen encodes an RNP motif protein that interacts with Hox pathways to repress the development of head-like sclerites in the Drosophila trunk

The homeotic genes in Drosophila can be divided into three classes: (1) genes of the homeotic selector, or Hox class, which are located in the Antennapedia and Bithorax complexes, (2) genes of the Polycomb (Pc) and trithorax (trx) classes, which regulate the spatial and temporal location of Hox transcription, (3)genes of the Hox modulator class, such as extradenticle, which act in parallel with Hox genes to regulate downstream targets. Hox genes are activated in distinct metameres on the anterior-posterior body axis and morphologically diversify those serially arranged fields of cells (Lewis, 1978; McGinnis and Krumlauf, 1992). The Hox proteins do this by their differential regulation of batteries of downstream ‘realizator’ genes (Garcia-Bellido, 1977; Pradel and White, 1998). Despite much research, the basis of differential realizator gene expression by Hox proteins is illuminated by only a few facts of unknown generality (Chan et al., 1997; Biggin and McGinnis, 1997; Pradel and White, 1998; Li and McGinnis, 1999; Li et al., 1999), and the basis of morphological differentiation under Hox control is even more obscure.

The ability of Hox proteins to differentially regulate batteries of downstream genes and thereby activate different developmental pathways requires a variety of modulatory factors that contribute to the binding occupancy and/or activity of Hox proteins on different target sequences. Such factors include the Hox proteins themselves, which can phenotypically suppress the activity of other family members, usually the posterior Hox function dominating the more anterior (Gonzalez-Reyes et al., 1990). Recent work by Capovilla and Botas (1998) suggests that this phenotypic suppression involves a short range enhancer repression mechanism (Gray et al., 1999). A few other Drosophila modulatory factors are apparently also involved in the specificity of Hox output, as evidenced by their homeotic mutant phenotypes in a wild-type genetic background. These include the DNA binding proteins extradenticle (exd) (Peifer and Weischaus, 1990; Rauskolb et al., 1993), teashirt (tsh) (Fasano et al., 1991; Alexandre et al., 1996), cap’n’collar B (cncB) (Mohler et al., 1995; McGinnis et al., 1998) and homothorax (hth) (Rieckhof et al., 1997; Pai et al., 1998).

Exd and its mammalian homologs in the Pbx class are homeodomain proteins (Flegel et al., 1993; Rauskolb et al., 1993) which participate in the selection and regulation of many Hox targets by cooperative binding with Hox proteins and concerted activity with them once bound (van Dijk and Murre, 1994; Chan and Mann, 1996; Mann and Chan, 1996; Chan et al., 1996; Pinsonneault et al., 1997; Li et al., 1999). The ability of Exd protein to enter nuclei and interact with Hox proteins is dependent on Hth, another Drosophila homeodomain protein (Rieckhof et al., 1997; Kurant et al., 1998; Pai et al., 1998). This activity of Hth, in combination with its pattern of expression, generates cellular patterns of nuclei that either contain or lack Exd protein. In cells that accumulate Exd in nuclei, Hox proteins are capable of serving as transcriptional activators, but in cells that lack nuclear Exd, Hox proteins have only been observed to function as repressors (Pinsonneault et al., 1997; Li et al., 1999). Tsh is a zinc-finger DNA binding protein, and is required in combination with the trunk HOX proteins to repress head-specific developmental pathways and promote trunk identity (Fasano et al., 1991; Röder et al., 1992; Alexandre et al., 1996). For example, Tsh acts with Antp to repress the ectopic activation of the anterior-most Hox gene, labial (lab; Röder et al., 1992). Finally, the CncB protein, a bZIP class of transcription factor, functions as a Hox modulator by preventing the Deformed Hox protein from activating maxillary-specific targets in the mandibular segment (Mohler et al., 1995; McGinnis et al., 1998).

Using genetic modifier screens, we have identified additional factors which influence the function of the Deformed (Dfd) gene in Drosophila, focusing on those that act in parallel to Dfd and other Hox genes (Harding et al., 1995; Gellon et al., 1997; Florence and McGinnis, 1998). The normal function of Dfd is required for the morphogenesis of structures arising from the mandibular and maxillary head segments, dorsal ridge, optic lobe and central nervous system (McGinnis et al., 1990; Restifo and Merrill, 1994). Our modifier screens identified mutant alleles of both exd and cncB, but also identified a gene with a particularly interesting mutant phenotype in which the ventral regions of thoracic (and occasionally abdominal) segments are homeotically transformed to head-like skeletal structures. We previously mapped this gene to the tip of the left arm of chromosome 2, at 21B4-6, and called it polycephalon (Gellon et al., 1997). However, we recently found that our polycephalon alleles disrupt the previously named split ends (spen) gene (Kolodziej et al., 1995), which was originally mapped to the right arm of the 2^nd chromosome. We will hereafter use the name spen to denote the gene. Mutant alleles in spen [called E(Raf)2A] also were isolated as enhancers of a dominant Raf-induced eye phenotype by Dickson et al. (1996), and yet other alleles of spen [called E(E2F)2A or poc) were isolated as enhancers of an eye phenotype induced by overexpression of dE2F and dDp (Staehling-Hampton et al., 1999).

The homeotic phenotypes in spen mutant embryos are not associated with changes in Hox gene expression, nor is the spen expression pattern changed in Hox mutants, therefore spen appears to act in parallel with Hox segmental identity functions. In addition, over-expression of the thoracic determinants Antp or tsh can suppress the ectopic sclerites in spen mutants, and thus we propose that spen acts in combination with Antp and Tsh pathways to regulate the distinction between head and trunk identity. The spen gene encodes very large protein isoforms (approx. 5,500 amino acids) with an arrangement of RNP motifs and conserved C-terminal regions that can be used to identify Spen orthologs and paralogs in D. melanogaster, C. elegans and H. sapiens.

As members of the RNP motif family exert their diverse functions by sequence-specific binding to RNA (Burd and Dreyfuss, 1994), the Spen proteins may modulate Hox function through a regulatory influence on messenger RNAs produced from Hox downstream genes.

Most mutations, aberrations and abbreviations are described in Lindsley and Zimm, (1992), or in FlyBase (http://flybase.bio.indiana.edu). Other alleles and lines include: spen^poc361 and spen^poc231 (Gellon et al., 1997), spen^E(Raf)2A16H1 and spen^E(Raf)2A1401 (Dickson et al., 1996; these are listed as polycephalon alleles in the current FlyBase), spen^E(CycE)D57 and spen^E(CycE)e9 alleles (Christian Lehner, personal communication), Df(2L)TE21A, Dll⁵, cnc^VL110, lab¹, oc², ems³, btd¹, Antp²⁵, Antp^hs.PHTA, Scr^C1Antp^NsRC3, Dfd¹⁶, tsh⁸, tsh^Scer/UAS.cGa, 69B-GAL4 (Brand and Perrimon, 1993; Chou et al., 1993), P{ry^+t7.2=neoFRT}40-A (‘FRT-40A’), P{w^+mC=ovoD1-18}2L1P{w^+mC=ovoD1-18}2L2P{ry^+t7.2=neoFRT} 40-A (‘ovo^D1-FRT-2L’), P{ry^+7.2=hsFLP}12; CyO/noc^Sco. The PZ1295 line [also known as l(2)03350)] was obtained from the Bloomington Stock Center. The l(2)K06805, l(2)K07612, l(2)K07721, l(2)K08102, l(2)K10325 and l(2)K13625 lines were obtained from the Berkeley Drosophila Genome Project (http://www.fruitfly.org). Embryonic stages are as described by Campos-Ortega and Hartenstein (1985).

Mutant chromosomes were outcrossed to eliminate balancer chromosomes from the stocks before mutant cuticles were collected. Embryos were collected for approximately 12 hours and aged for more than 24 hours before preparing cuticles as described by Wieschaus and Nüsslein-Volhard (1986). Overexpression of tsh in a spen mutant background was accomplished by crossing spen^poc361/CyO;UAStsh/+ females to spen^poc361/+;69B-GAL4/+ males. Overexpression of Antp in a spen mutant background was accomplished by crossing spen^poc361/+;hsAntp/+ females to Df(2L)TE21A/+;hsAntp/+ males, collecting embryos for 1 hour intervals, aging at 25°C for 3 hours, and then heat shocking at 37°C for 20 minutes every 2 hours for a total of three heat treatments. Embryos were then aged at 25°C for at least an additional 22 hours.

Maternal germline spen mutant clones were generated using the FLP-FRT system as detailed by Chou et al. (1993). spen^poc231 and spen^poc361 were recombined onto the FRT 40A chromosome. spen FRT 40A/CyO virgin females were then mated to hs-FLP12; ovo^D1-FRT-2L/CyO males; progeny of this cross were heat shocked for 1 hour at 37°C at 3, 4, 5 and 6 days after egg deposition. The heat-shocked female progeny carrying both the spen and ovo^D1 chromosomes were mated to males heterozygous for the other spen allele (e.g. spen^poc361/ovo^D1 females were mated to spen^poc231/CyO,wg males).

Genomic DNA from the PZ1295 line corresponding to l(2)03350 was prepared by standard protocols as described in Gellon et al. (1997). Rescue of the DNA surrounding the P-element insertion site was accomplished by digestion, ligation and transformation of the genomic DNA as described in Gellon et al. (1997).

Excision lines were generated from PZ1295 by using the Δ2-3 strain to mobilize the element (Robertson et al., 1988). Genomic DNA was prepared from homozygous viable and homozygous lethal excision lines and mapped by Southern analysis using probes surrounding the region obtained by plasmid rescue. A similar analysis was subsequently done to map the P insertion sites in the l(2)K06805, l(2)K07612, l(2)K07721, l(2)K08102, l(2)K10325 and l(2)K13625 alleles.

We initiated a walk in the region surrounding the insertion point of PZ1295 using a 3.2 kb HindIII/XbaI genomic fragment obtained by plasmid rescue. Approximately 50 kb of overlapping λ clones were obtained from an EMBL-3 iso-1 genomic library provided by J. Tamkun using standard protocols and nick translated probes. An additional 30 kb proximal to our walk was kindly provided by M. Noll (λ clones y17-3, y16-1, y15-1).

Poly(A)⁺ RNA was prepared from 0-to 2-hour, 2-to 8-hour, 8 to 12-hour and 12-to 24-hour embryos, purified once over an oligo-dT column, electrophoresed in a formaldehyde gel and blotted according to standard procedures. Northern blots were hybridized with ³²P-labeled DNA probes.

cDNAs were isolated from an oligo-T primed 8-to 12-hour embryonic cDNA library provided by Nick Brown (Brown and Kafatos, 1988), and a randomly primed 0-to 18-hour embryonic λGT10 cDNA library (Clontech). Probes were labeled by nick translation. Hybridization and washes were done at high stringency following standard protocols. More than 100 overlapping partial cDNAs corresponding to the putative spen transcript were isolated using genomic and cDNA fragments as ³²P-labeled probes from the region shown in Fig. 2. Thirty cDNA clones were subcloned and mapped.

A short region (150 bp) within the 4.8 kb BamHI-EcoRI fragment that spans most of the 3′ half of the large exon of the putative spen transcript sequence was not included in any of the cDNA clones. RT-PCR was used to determine that this sequence is included in spen transcripts. Poly(A)⁺ RNA, isolated as described above, was used as a template for reverse transcriptase following the protocols included with the enzyme, Superscript II Reverse Transcriptase (GibcoBRL). The reaction was primed with a sequence-specific 3′ primer. PCR was performed using the 1^st strand reaction as template and following the protocol included with the RT enzyme (GibcoBRL).

Sequencing was performed mainly at the Center for AIDS Research CORE Facility at UCSD. The BLAST algorithm (Altschul et al., 1997) was used to search protein and EST databases for sequences homologous to short fragments of protein sequence. The 3 RNP-type RNA binding domains were identified in this manner. Comparison of the Spen protein sequences (accession no. AF188205) to the ISREC PROSITE database identified five bipartite nuclear localization motifs. Searches of genomic and EST databases revealed cDNA and genomic sequences which are predicted to encode the protein domains shown in Fig. 4; from D. melanogaster (accession no. for short form –AAD38639), Homo sapiens Spen-like (accession no. for long form –AL034555, for short form – EST sequence AI621042, and genomic L13434), H. sapiens RNP2 from U1a protein (accession no. 2554638), Caenorhabditis elegans (accession no. for long form – CAA91320; for short form – AAC19192) and Nicotiana tabacum (accession no. U90212), Xenopus laevis (accession no. P20397).

RNAi was performed essentially as described by Kennerdell and Carthew (1998). A 972 bp DNA fragment from the region that encodes the 3′ UTR of the spen transcript (from 11 nucleotides to 983 nucleotides downstream of the Spen TAA stop codon) was transcribed with T7 RNA polymerase to generate RNA copies of both strands. The resulting double-stranded RNA was injected into the middle region of embryos less than one hour after egg deposition. Controls consisted of injected buffer at the same site and injection of cap’n’collar B (cncB) RNA. cncB loss-of-function mutants have an obvious homeotic phenotype in embryos (McGinnis et al., 1998) that is mimicked by injection of ds cncB-specific RNA (A.Veraksa and W. M., unpublished results). Injected embryos were aged and their cuticles prepared after removing the vitelline membranes by hand.

Point mutations in spen open reading frames were detected by denaturing gel electrophoresis of 400-500 base pair fragments of genomic DNA isolated from spen heterozygotes and parental strains. Genomic DNA was purified using the QIAamp Tissue Kit (QIAGEN), and PCR was performed using primers designed to produce fragments with the melting behavior described by Myers et al. (1987). Melting point variations resulting from point mutations in the fragments were detected using a Denaturing Gel Electrophoresis System (BioRad). Fragments which showed melting point variation were sequenced on both strands to determine the specific nucleotide position that was changed in mutant chromosomes.

A GST-Spen fusion protein containing amino acids 3203-3714 from the common region of the SpenL protein was injected into mice to generate anti-Spen antiserum. The fusion protein was bacterially expressed and purified using glutathione-Sepharose (Pharmacia) and standard techniques. For staining with this antibody, embryos were collected, dechorionated and fixed for 15 minutes in 4% formaldehyde. The antiserum was used at a 1:100 dilution, and detected with FITC-labeled anti-mouse secondary antibody (Jackson Labs).

In situ hybridizations to detect transcript expression were performed using a variation of the protocol described by Tautz and Pfeifle (1989). DIG-labeled antisense probes were made from the following cDNA sequences: Dfd probe was generated from a 2.3 kb EcoRV-SmaI fragment isolated from pcDfd41 (Regulski et al., 1985), cloned into the EcoRV site of pBluescript, linearized with BamHI and transcribed from the T3 promoter. Scr probe was generated from a 2.1 kb EcoRI-XbaI genomic fragment containing the homeobox (Kuroiwa et al., 1985), subcloned into pGEM2, linearized with EcoRI and transcribed from the T7 promoter. Antp probe was generated from a 2.2 kb BamHI-EcoRI fragments from p903 (Hafen et al., 1983) subcloned into pBluescript, linearized with BamHI and transcribed from the T7 promoter. spen antisense transcript probes were generated from a 5.2 kb cDNA fragment in clone pc1.65BH10-5.2R that included the RNP coding region subcloned into pBluescript, linearized with NotI and transcribed from the T3 promoter.

We previously reported the isolation of six EMS-induced mutant alleles of spen based on its interaction with the Hox gene Dfd (spen was called polycephalon by Gellon et al., 1997). Some of these alleles behave as genetic nulls for the cuticular phenotype, i.e. spen³⁶¹/spen³⁶¹ cuticles are indistinguishable from spen³⁶¹/Df(2L)TE21A(deletes the entire spen locus), and spen²³¹ is slightly weaker than spen³⁶¹. Loss of zygotic spen function results in embryonic lethality that is associated with a loss of the anterior portion of the H-piece, and with a kinked median tooth (Fig. 1B,F, compare to Fig. 1A). The anterior parts of the H-piece are derived from the ventral maxillary segment, and are dependent on Dfd function (Merrill et al., 1987; Regulski et al., 1987). Head development is profoundly disrupted in maternal/zygotic spen mutants. These mutant embryos have non-involuted heads that are missing many head sclerites, including the base of the mouth hooks, median tooth, anterior regions of the lateralgräten, and the dorsal bridge. The ventral arms and vertical plates are also strongly reduced (Fig. 1D,F,). Many of these spen-dependent sclerites are also dependent on head Hox genes such as labial (lab), and Sex combs reduced (Scr), gap/homeotic genes such as empty spiracles (ems) and orthodenticle (otd), and other head patterning genes (Merrill et al., 1989; Sato et al., 1985; Pattatucci et al., 1991; Dalton et al., 1989; Finkelstein and Perrimon, 1991). In the maternal/ zygotic mutants, the trunk region of the embryos appears to be largely unaffected. The denticle belts and posterior structures such as filzkörper appear normally shaped, although these structures, as well as the remnants of head skeleton (and ecopic head-like sclerites, see below), are somewhat less pigmented/ sclerotized than normal. The overall low level of sclerotization may be due to embryonic death in maternal/zygotic mutants before the cuticle is fully developed (Fig. 1D).

Approximately 50% of the embryos that are zygotic or maternal/zygotic spen mutants also develop sclerites in the thoracic segments. These sclerites appear most frequently in the second and third thoracic segments (T2 and T3), and overlap the anterior/posterior compartment boundary. The expressivity of this phenotype ranges widely. The weakest phenotypes exhibit a row of small chunks of ectopic sclerotic material, usually limited to T2 and T3 (Fig. 1B). Strong phenotypes show broad bands of sclerotic material in all thoracic segments, often accompanied by sclerotic patches in lateral regions of the abdominal segments (Fig. 1C). No matter how extensive, the sclerites do not form in the fields of denticle belts, nor do they change the overall size of a segment. In small patches, the ectopic material is variably sclerotized, scar-like, and not recognizable as any other embryonic structure. When manifest in broad patches, the sclerotic material is brown and striated, reminiscent of the brown, striated appearance of the ventral arms, vertical plates and dorsal arms of the head skeleton (Fig. 1E). We conclude that one role of spen is to suppress the production of head-type sclerotization from ventral thoracic cells, and it is needed to a lesser extent for this function in lateral abdominal cells.

This spen⁻ homeotic phenotype is not specific to alleles isolated on the basis of Dfd interaction. In a screen for suppressors and enhancers of an activated Raf construct, Dickson et al. (1996) isolated three alleles of E(Raf)2A. These alleles do not complement the lethality of our spen alleles, and we find that the E(Raf)2A alleles also exhibit ectopic head-like sclerites, as do 11 additional alleles of spen that enhance the rough eye phenotype generated by ectopic expression of Cyclin E in the eye (Christian Lehner, personal communication).

Production of ectopic head-like sclerotic material in the trunk indicates that loss of spen function might result in de-repression of head patterning genes. We examined the cuticular phenotypes of embryos doubly mutant for spen and Distal-less (Dll), cnc, lab, Dfd, buttonhead, otd and ems, all of which have been shown to be required for determining head-specific pathways (Mohler et al., 1995; Cohen and Jurgens, 1990; Finkelstein and Perrimon, 1991). All of the double mutant combinations show the same degree of thoracic sclerotization as spen single mutants. We also examined the transcription patterns of Dll, cnc, lab, Dfd, ems, otd and apontic in spen mutant embryos and found them to be indistinguishable from wild type (data not shown). Therefore we conclude that spen does not repress ectopic head-like cuticle by repressing the expression or function of these known head-determining genes. Finally, ectopic expression of Dfd and lab in spen mutant embryos did not alter the degree or character of the ectopic sclerites in the thorax (data not shown).

To understand how mutations in spen produce ectopic sclerites, we cloned the gene which encodes the spen function. All six of the EMS-induced spen mutant alleles recovered from the Dfd modifier screen (Gellon et al., 1997) fail to complement the lethality of the P-element insertion PZ1295 (l(2)03350), located within cytological interval 21B4-6. Precise excisions of the PZ1295 insertion are viable over all six spen alleles, whereas imprecise excision lines are still lethal over spen mutant chromosomes. These results suggest that PZ1295 interrupts a sequence crucial to the function of the spen locus. Genomic DNA adjacent to the site of PZ1295 insertion was rescued and overlapping λ clones spanning approximately 50 kb of surrounding genomic DNA were isolated. In addition, genomic clones covering the 3′ end of what turned out to be the spen transcription unit were kindly provided by Markus Noll. A structural map of the locus is shown in Fig. 2A. Northern analysis with genomic probes detected two transcripts from the region, one of 1.3 kb (X in Fig. 2A) and another of 20 kb (spen in Fig. 2A). The 3′ end of kismet (kis) is located 7.5 kb from the 3′ end of spen (Fig. 2A; Daubresse et al., 1999).

Five of the spen⁻ P-element insertions map either in the first exon or the first intron of the long isoform of the 20 kb transcript, while one maps immediately 5′ of this transcript (Fig. 2A). Evidence that the 20 kb transcript corresponds to the spen locus is based on: (a) reversion of the mutant phenotype upon excision of the P element inserts; (b) identification of amino acid changes in two EMS-induced spen alleles, both isolated as enhancers of CycE (Christian Lehner, personal communication; see Fig. 4); and (c) the generation of spen⁻ head defects and ectopic head-like sclerites upon injection of double stranded RNA specific to the 20 kb transcript sequence (data not shown).

Using genomic DNA fragments from these regions, approximately 200 cDNA clones were isolated from a poly-T primed, 8-12 hour embryonic cDNA library (Brown and Kafatos, 1988), and from a randomly primed, 0-18 hour embryonic cDNA library. Overlapping cDNA clones were mapped and sequenced, resulting in the exon/intron structure shown in Fig. 2A. The structure of the cDNA clones that we analyzed suggests that at least two alternative promoters initiate transcripts. A probe made from the most 5′ exons, which initiate the long (spen^L) transcript isoform does not detect maternal transcripts in the 0-2 hour RNA lane of a northern blot (Fig. 2B, probe 1), whereas a probe from the common region does (Fig. 2B, probe 2). However, RT-PCR analysis (not shown) does detect spen^L and spen^S transcripts at all stages, including 0-2 hours, indicating that low levels of the spen^L isoform exist in maternal RNA. Both probes detect transcripts of approximately 20 kb at later stages of embryonic development, although the spen^L isoform is predicted to be approximately 1 kb larger based on the known cDNA structures. A third Spen protein isoform has been characterized by B. Kuang, S. Wu, Y. Shin, K. A. Mace, L. Luo and P. Kolodziej (personal communication), and is initiated from an alternative exon that maps upstream of the spen^S 5′ exon. The third isoform has a few unique amino acid residues at its N terminus that differ from two Spen protein isoforms reported here.

The first ATG in the long open reading frame of the spen^Ltranscript (Fig. 2) begins a long open reading frame of 5533 codons, which is predicted to make a 597 kDa protein (Fig. 3). The first ATG in the spen^S transcript begins an open reading frame of 5476 codons. The SpenS protein isoform is essentially an N-terminal truncation of the SpenL protein with the addition of a Met Arg dipeptide, encoded in S exon 2, substituted for the first 59 aa residues of SpenL (Fig. 3). The N terminus of the SpenL isoform may contain unique functions due to its poly-Asn and poly-Gln sequences (Fig. 3). Three RNA-binding domains of the RNP type (amino acids 554 – 806 of L) are present in the 5474 aa common region of the predicted SpenL and SpenS proteins. Five bipartite nuclear localization sequences are scattered throughout the middle of the common sequence (positions 1872-1889, 1949-1966, 2135-2152, 2450-2467, 4556-4573 of SpenL). In addition, a region that is highly enriched in Glu (E) and Lys (K) residues is present from positions 1917-2053 of SpenL, and strings of glutamine (Q) residues are found throughout the C-terminal 1/3^rd of the Spen proteins (Fig. 4A).

Database searches combined with GENESCAN and FGENE predictions of exon structure reveal apparent structural orthologs of Spen in genome sequences from human and C. elegans. These Spen family members are defined by their large size (predicted to be at least 3300 aa for human and 2738 aa for C. elegans), by their possession of three RNP domains that are closely related to those in Spen isoforms, and by a lengthy sequence match to a 165 amino acid motif at the C terminus of the D. melanogaster Spen proteins (Fig. 4A). The sequence of the first RNP domain of the Spen family (RNPa in Fig. 4B) also is closely related to the first RNP domain of Nucleolin, a protein involved in ribosome biosynthesis (Ginisty et al., 1999). The second Spen family RNP domain (RNPb in Fig. 4B) is a good match to the canonical RNP domain (Fig. 4B, bottom line), and shares even more identities with an RNP domain in AC binding factor (ACBF), an RNP-motif protein isolated from tobacco. ACBF binds to DNA regulatory sequences of genes in the phenylpropanoid biosynthetic pathways of many plants (Séguin et al., 1997).

The C-terminal motif, which we call the SPOC domain (Spen Paralog and Ortholog C-terminal domain) is very similar among family members (57% identity between D. melanogaster Spen and H. sapiens Spen1; Fig. 4C), but matches no other sequence motifs with known biochemical function. The entire open reading frames of two EMS-induced spen alleles (spen^E(CycE++)D57 and spen^E(CycE++)e9) were sequenced. Single base substitutions in each of these alter the predicted amino acid sequence at highly conserved positions within the SPOC domain: codon position 5444 of spen^E(CycE++)e9, TGC – Cys to TAC – Tyr; and codon position 5480 of spen^E(CycE++)D57, GGC – Gly to GAC – Asp (Fig. 4C). With regard to the embryonic head and thoracic defects observed in spen³⁶¹ zygotic mutants, both of these alleles are hypomorphs. The order of allele strengths with respect to zygotic defects is spen³⁶¹ > spen²³¹ > spen^E(CycE++)e9 > spenE(CycE++)D57.

Our sequence similarity searches also reveal a separate sub-family of short Spen-like proteins encoded in D. melanogaster and other animal genomes. The short Spen-like proteins have a similar arrangement of Spen-like RNP motifs and SPOC domains to those in the Spen orthologs, but the short Spen-like proteins are approximately one tenth the size of Spen (Fig. 4). All of the known short Spen-like proteins also encode an RGG motif, which is capable of destablizing RNA helices and often found in combination with RNP motifs (Siomi and Dreyfuss, 1997). The current sequence evidence suggests that a duplication event giving rise to the genes that encode the large and small Spen-like proteins occurred before the divergence of the deuterostome and protostome lineages in metazoan evolution (Aguinaldo et al., 1997).

Antibodies were raised against a common fragment of the Spen proteins (amino acids 3203-3714) to assay the expression pattern and subcellular localization. Immunostaining of whole-mount embryos shows that Spen antigen is expressed in most or all cell types, and is concentrated in nuclei (Fig. 5A-C). Spen protein staining is first detectable prior to cellularization in stage 3 embryonic nuclei, and is present in all blastoderm cells, including the pole cells (Fig. 5A). The protein is expressed throughout the rest of embryogenesis, and is concentrated in nuclei. At later embryonic stages (9 through 14), nuclear Spen staining appears to be most abundant in CNS and epidermal cells (Fig. 5C), but most tissues exhibit nuclear staining at detectable levels.

Null mutations in Antp result in a transformation of T2 and T3 towards T1 in the embryonic body plan (Wakimoto and Kaufman, 1981; Martinez-Arias, 1986). In addition, Antp mutant embryos develop ectopic head-like sclerites in the dorsal thorax (between T1 and T2; Fig. 6A), similar in kind but not in position to the ectopic sclerite phenotype seen in spen mutants. To test whether spen and Antp function in an additive or synergistic manner in the repression of head-like sclerites in the thorax, we first studied spen⁻; Antp⁻ cuticle phenotypes. Embryos mutant for both spen and Antp have more sclerotic material in dorsal T2 than do Antp mutants alone (Fig. 6B). In addition, the ectopic head-like sclerites in the ventral thorax of spen⁻; Antp⁻ mutants are more sclerotized and extensive than in spen mutants alone (Fig. 6D, compare to 6E). The sclerotic material in spen⁻; Antp⁻ mutants frequently appears in two distinct bands, one in the center of the segment similar to the position in spen mutants, and at another position in the posterior of T1 and T2 (Fig. 6D, arrow). These posterior ectopic sclerites do not develop in T3 (Fig. 6D). The enhanced formation of head-like sclerites in spen⁻; Antp⁻ mutants suggests that spen and Antp function in a common or interacting pathway(s) in subregions of T1 and T2.

The synergistic effect of Antp and spen might be due to a regulatory effect of Antp on spen transcription pattern, or to Spen effects on Antp transcript pattern or translation. However, Antp transcript and protein expression patterns are unchanged in spen mutant embryos, and spen transcript expression is unchanged in Antp mutant embryos (data not shown). Therefore, spen and Antp appear to be acting in parallel, presumably due to direct or indirect regulation of common downstream genes.

If spen and Antp regulate common targets, then induction of high levels of exogenous Antp expression might result in suppression of the spen mutant phentoype. We therefore tested the ability of excess Antp protein to suppress the spen mutant phenotype. Overexpression of Antp under heat shock promoter control (hsAntp) causes a transformation of head regions to thoracic identity, but leaves T2 and T3 nearly unchanged (Fig. 6F; Gibson and Gehring, 1988). When Antp is overexpressed in a spen mutant background, the ectopic head-like sclerites are strongly suppressed (Fig. 6G). The number of hsAntp, spen⁻ embryos which exhibit any detectable ectopic sclerites was less than half the expected number compared to spen⁻ mutant siblings from the same cross, or compared to spen⁻, hsAntp embryos that were not subjected to heat shock. In addition, the sclerites which do occasionally appear in heat shocked hsAntp; spen⁻ embryos are smaller than those in their spen⁻ siblings (Fig. 6G,H) The ability of excess Antp to suppress the spen⁻ homeotic transformation indicates that the two genes interact to repress ectopic head-like sclerites.

In the head region, where spen is required for the development of sclerites that also require Dfd, Scr and other head genes, it is also possible that spen might work in parallel to Hox pathways. This is the case for Dfd and Scr, since mutations in these genes had no effect on spen transcript expression pattern, and conversely, spen mutants had no effect on Dfd or Scr transcript or protein expression patterns (data not shown). We attempted to test whether the overexpression of Dfd could rescue the H-piece defect in spen mutant embryos, but the morphology of the head skeleton was so disrupted by heat shock induced ectopic Dfd protein (Kuziora and McGinnis, 1988) that it was not possible to conclude whether the anterior H-piece was restored.

Another Drosophila gene involved in distinguishing head from body is teashirt (tsh). Tsh protein is expressed only in the labial segment and trunk region of embryos, where it is required to repress head identity and to promote thoracic and abdominal segment identities (Fasano et al., 1991). tsh transcription levels in the thorax are maintained by Antp (Röder et al., 1992; McCormick et al., 1995), but a variety of genetic interaction tests have shown that Antp and Tsh have independent functions in repressing head development (Röder et al., 1992). Is spen integrated into the Antp; tsh pathways by regulation of the tsh or Antp expression patterns. Our experiments show that: (1) expression pattern of Tsh protein (Röder et al., 1992) is unchanged in spen mutant embryos, (2) protein expression patterns of Scr or Antp in spen⁻, tsh⁻ double mutant embryos are unchanged from the pattern seen in tsh mutants alone (data not shown), and (3) the spen transcript pattern is normal in tsh mutants (data not shown). Therefore, Spen suppression of head-like sclerites is not exerted by a regulatory effect on the Tsh protein expression pattern, nor by Tsh effects on the spen transcript pattern, nor through combinatorial effects of spen and tsh on Scr or Antp protein abundance.

The phenotype of tsh⁻, spen⁻ mutant embryos suggests that the two genes act to promote thoracic development. In tsh mutant embryos, the T1 denticle belt is absent and although the remaining denticle belts appear to have the appropriate segmental identities, the denticles themselves are disorganized and smaller than in wild type (Fig. 7A; Fasano et al., 1991; Röder et al., 1992). In contrast, tsh⁻, spen⁻ double mutants completely lack denticle belts in the thorax (Fig. 7B). This may be due to the the death of cells in the denticle field in the thorax of the double mutants, or to the inability of Antp protein, still expressed in the remaining cells, to promote the development of thorax-specific structures.

As to whether tsh and spen collaborate in repressing head-like sclerites, we find that the tsh⁻, spen⁻ double mutants still have bits of sclerite in the ‘thorax’ of the double mutants, so this phenotype is not enhanced. However,we also examined the effects of Tsh overexpression on the ectopic head-like sclerites in spen mutants. In wild-type embryos, overexpression of Tsh protein throughout the embryo results in transformation of head regions toward thoracic identity (de Zulueta et al., 1994), as well as poorly differentiated denticle belts, especially in the thorax. In the thorax of spen mutant embryos that also overexpress Tsh, the ectopic ventral head-like sclerites are strongly suppressed (Fig. 7D). Taken together, these results suggest that spen, tsh and Antp function in a combinatorial manner to repress the development of head-like sclerites and promote the development of thoracic identity.

Encoding RNA binding domains, spen is a unique type of homeotic function. It is neither in the Hox class, nor is it required for the proper spatial regulation of Hox transcription, such as trx or Pc group homeotics. It falls into the growing class of functions that act in parallel to modulate the output of Hox pathways. The direction of spen’s modulatory effect, either helping to promote or to repress head-type sclerotization, is apparently dictated by the type of Hox gene expressed in a certain region of the body.

Although expressed throughout the embryo, spen appears to be especially important to the morphogenesis of head cuticular structures, presumably due to modulation of its function on the anterior/posterior axis by specific Hox proteins and specific signaling pathways. The epistasis experiments reported here, along with previous results, indicate that in its role of promoting sclerite formation in the head, spen acts in parallel to head patterning genes such as Dfd and ems (Fig. 8). There are also other embryonic head structures that are missing or abnormal in both maternal/zygotic spen mutants and other head homeotic mutants such as lab and Scr (Merrill et al., 1989; Sato et al., 1985; Pattatucci et al., 1991). The homeotic function of spen is exerted in the trunk, where it represses head-like sclerite formation in combination with both the segment identity homeotic Antp and the regional homeotic gene tsh (Fig. 8).

The thorax-promoting spen function also genetically interacts with both tsh and Antp. In double mutants for spen and tsh, the thorax has few segmental specializations. This is an interesting result, for Antp expression is unaffected in the epidermal cells that remain in the thoracic region. Another result indicating a close genetic interaction between tsh and spen is that overexpression of Tsh protein can suppress the ectopic sclerite phenotype of spen mutants. The Spen and Tsh functions, both of which act in parallel to the Hox proteins, seem likely to exert this effect through by their collaboration (direct or indirect) on the regulation of Hox downstream target genes. Presumably this occurs by the Tsh DNA binding protein operating as a transcription factor (Alexandre et al., 1996), and by Spen acting at the level of RNA processing/transport.

Staehling-Hampton et al. (1999) recently discovered that in Drosophila eye cells, wild-type spen genetically antagonizes ectopically expressed E2F and Dp proteins, and genetically assists p21 in blocking S-phase entry. Thus at least in the eye, spen apparently has a role in suppressing progression of the cell cycle. Other evidence consistent with this derives from the study of interactions between spen and Raf mutations. spen^E(Raf)2A alleles were found to enhance a dominant, constitutively active Raf phenotype in the R7 photoreceptors of the Drosophila eye (Dickson et al., 1996). Clones of spen mutant cells in the eye paradoxically exhibited either loss of R7 cells, or supernumerary R7 and other photoreceptors (Dickson et al., 1996). Given the interactions and phenotypes they observed, Dickson et al. (1996) concluded that at least in R7 cells, spen acts downstream of Raf, as a constitutive repressor of Raf signaling, in a manner similar to the ETS transcription factor Aop/Yan (Nüsslein-Volhard et al., 1984; Lai and Rubin, 1992). It is interesting to note that the maternal/zygotic loss of function mutations reported here for spen, and those for null zygotic mutations of aop/yan (Rogge et al., 1995) are similar in that thoracic and abdominal features are relatively normal, while head skeletal structures are missing or malformed. So it is possible that Spen may act along with Aop/Yan in the nuclei of embryonic head cells as a repressor of Ras/Raf signaling.

Genetic interactions between Ras signaling and Hox function have been described previously; Boube et al. (1997) found that proboscipedia (pb) loss-of-function homeotic phenotypes in the Drosophila labium were enhanced in Ras1⁻/Ras1⁺ adults, and Ultrabithorax (Ubx) loss-of-function homeotic phenotypes were enhanced when haltere cells contained only one wild-type allele of Gap1 (a Ras inhibitor). In C. elegans, the ability of the lin-39 Hox gene to specify vulval cell fates requires activation of the Ras pathway (Clandinin et al., 1997; Maloof and Kenyon, 1998; Eisenmann et al., 1998). Ras acts both in parallel to lin-39 as a co-activator of vulval identity, and also amplifies the amount of lin-39 protein expression in vulval precursor cells. Another C. elegans Hox gene, egl-5, also requires Ras signaling to provide a neuroectoblast cell fate to the P12 cell, which is accomplished at least in part by Ras pathway activation of the transcription of egl-5 (Jiang and Sternberg, 1998). Thus the impact of Ras/Raf signaling on Hox gene expression and Hox protein function may be commonly used in developmental genetic circuitry. In light of all the genetic interactions of Spen, it is possible that it has a general function in inhibiting cell division and acts in combination with Hox, E2F, and Ras/Raf signaling pathways to accomplish this role. How this putative general growth control function might be connected (if at all) to the loss of sclerites in the head and the ectopic production of head-like sclerites in the trunk is unknown.

The extent to which Spen connects Ras/Raf pathways and Hox pathways is a fascinating but still largely unexplored area, as are the mechanisms by which Spen fits into these pathways. RNA-binding proteins have been postulated to function in signaling pathways by binding signal-responsive gene transcripts (Siomi and Dreyfuss, 1997). In this postulated mechanism, a transcript might be released for translation in response to activation of a signaling pathway, providing very rapid protein induction; or a different protein isoform might be generated by differential splicing in response to signaling pathways. However, these and other possible mechanisms are speculative until common target genes of Hox and Spen, or Spen and Raf, are identified.

The ectopic head-like sclerites which develop in the thorax of spen mutant embryos are darker, more striated, and thicker than normal thoracic body wall cuticle. In addition, spen is required for the formation and sclerotization of head structures such as the H-piece and ventral arms. Thus a common theme of spen function in the epidermis is regulation of sclerotization. What are possible ‘sclerite’ target genes that might be activated by Dfd and Spen, and repressed by Antp and Spen?

At the molecular level, the formation of sclerites involves the incorporation of a complex array of substances in addition to chitin. Among these substances are cross-linking agents that generate darkly pigmented, thick, and hard sclerites (Poodry, 1980). Therefore, in thoracic cells, Spen and Antp are directly or indirectly repressing the expression or activities of the enzyme and transport pathways that synthesize and secrete these sclerotization-promoting substances. Genes in the pathways that synthesize some of these substances have been long studied in Drosophila, and include Dopa decarboxylase (Ddc) and pale (tyrosine hydroxylase; Wright, 1996; Jurgens et al., 1984; Neckameyer and White, 1993). These enzymes are required for the production of dopamine, which functions both as a neurotransmitter in the fly nervous system, and a precursor of cross-linking molecules required for the sclerotization of cuticle in the body wall. It will be interesting to investigate whether Ddc, pale, or other genes in the dopamine pathway are targets of the repressive functions of spen and Antp in thoracic cells.

Spen and its homologs contain three clustered RNP-type RNA binding domains (Siomi and Dreyfuss, 1997; Handa et al., 1999). Other proteins which contain RNP domains include Sex lethal (Sxl), U1A snRNP, Nucleolin, and the circadian rhythm regulator Lark (Burd and Dreyfuss, 1994; Serin et al., 1997; Newby and Jackson, 1996). It therefore seems highly likely that Spen family protein functions will be exerted by RNA binding.

We thank T. Kaufman, J. Mahaffey, G.Morata, G. Struhl, S. Kerridge, Christian Lehner, Barry Dickson and Ernst Hafen for providing fly stocks, J. Tamkun and N. Brown for sharing their Drosophila libraries. Christian Lehner and Peter Kolodziej generously shared unpublished results. We are also grateful to M. Noll for providing genomic clones in the spen region, Eileen Furlong and Matt Scott for anti-Antp and anti-tsh antisera, and J. Tamkun for sharing the northern analysis of the kis region prior to publication. Both FlyBase and the Berkeley Drosophila Genome Project were of great value at many stages in the completion of these experiments. This work was supported by NIH grant HD 28315.

The isolation of additional spen mutant alleles, along with the sequence of one of the Spen protein isoforms is also reported by Rebay, I., Chen, F., Hsaio, F., Kolodziej, P. A., Kuang, B. H., Laverty, T., Suh, C., Voas, M., Williams, A. and Rubin, G. M. (1999). A genetic screen for novel components of the Ras/MAPK signaling pathway that interact with the yan gene of Drosophila identifies split ends, a new RRM motif containing protein. Genetics, in press.