Cis-regulatory control of the SM50 gene, an early marker of skeletogenic lineage specification in the sea urchin embryo

Cis-regulatory target sites encode the recognition elements by which a gene interprets its spatial position in development. Here we demonstrate the individual functions of specific target site sequence elements that are required to generate accurate transcriptional expression of a gene utilized exclusively in the skeletogenic lineages of the sea urchin embryo. From a developmental point of view, the gene we have chosen for this study, SM50, provides an excellent molecular marker of the autonomous specification process by which the skeletogenic lineages are specified, as we discuss below. We discovered that a synthetic oligonucleotide representing a single short cis-regulatory site is capable of imposing skeletogenic lineagespecific expression on an entirely heterologous expression vector.

In regularly developing sea urchins, the skeletogenic lineages derive from four polar founder cells, the large micromeres. These segregate from their non-skeletogenic sister cells, the small micromeres, at fifth cleavage. Three further divisions of these lineages ensue during late cleavage, so that the embryo now contains 32 skeletogenic precursor cells. After gastrulation another division of these cells occurs so that the embryo ultimately contains 64 skeletogenic mesenchyme cells (Cameron et al., 1987; Ruffins and Ettensohn, 1993). At the early blastula stage, the skeletogenic precursor cells are located symmetrically in the vegetal wall of the embryo, surrounding the descendants of the small micromeres. They are surrounded by the vegetal plate precursors of the archenteron, with which they are in close contact. Midway into blastulation the skeletogenic precursors undergo a stereotypic differentiation process, as a result of which they acquire mobility, alter their cell surface constituents and properties, and reorganize their cytoskeletal structure. They then ingress singly into the blastocoel and assume a mesenchymal character (for review see Davidson, 1986; Ettensohn and Ingersoll, 1992; McClay et al., 1992). At the early gastrula stage the skeletogenic mesenchyme cells aggregate around the base of the invaginating archenteron. Skeletal elements, or spicules, then begin to be secreted, in two bilateral clusters of skeletogenic cells that form on the oral side of the embryo. Many genes that are specific to the skeletogenic process have been cloned (cf. reviews cited above; also Drager et al., 1989). At present the best known of these genes are msp130, which encodes a cell surface glycoprotein (Leaf et al., 1987; Anstrom et al., 1987; Harkey et al., 1992; Kabakoff et al., 1992), and the SM30 (George et al., 1991; Akasaka et al., 1994) and SM50 genes (Benson et al., 1987; Sucov et al., 1987, 1988; Katoh-Fukui et al., 1991). Both SM30 and SM50 encode proteins of the matrix within which the mineral elements of the skeletal structures are embedded. SM50 is of particular interest as a marker of the initial process by which the skeletogenic lineages are specified. Its transcripts begin to accumulate in the skeletogenic precursor cells within only two to three cleavages after the definitive 5th cleavage segregation of these lineages (Benson et al., 1987; Killian and Wilt, 1989). This is almost a whole day before the process of skeletogenesis per se begins to take place in the species with which this work was carried out, Strongylocentrotus purpuratus, and is also far in advance of any of the blastula-stage differentiation events that lead to ingression and expression of mesenchymal functions. We regard SM50 gene expression as an early marker of skeletogenic lineage specification.

Specification of the skeletogenic founder cells in regularly developing sea urchins is, by all available tests, an autonomous process (Davidson, 1989, 1990; McClay et al., 1992). No immediate intercellular interactions are required for this specification to occur. Thus, once the skeletogenic micromeres have formed, they can be transplanted to ectopic locations in the embryo, whence their progeny ingress and carry out skeletogenesis, irrespective of their new location (HÖrstadius, 1939; reviewed by Davidson, 1989). These cells are even unable to produce progeny that express any other differentiated fates, in contrast to all other blastomeres of the cleavage-stage sea urchin embryo (excepting their small micromere sister cells). Furthermore, as first shown by Okazaki (1975), if placed in culture, isolated 4th cleavage micromeres will carry out skeletogenesis in vitro, after undergoing the stereotypic number of divisions (reviewed by Davidson, 1986, p. 221); the only requirement is the addition of a small amount of horse serum to the sea water. The micromeres that harbor the autonomous potential to give rise to the skeletogenic precursor lineages are formed by the horizontal 4th cleavage plane, which is positioned so that they together include only the polar 7.5% of the cytoplasmic volume of the egg (Ernst et al., 1980). The implication of this autonomous specification process (Davidson, 1989, 1990) is that the activation of the earliest cohort of genes expressed exclusively in the skeletogenic lineage might depend on maternal factors localized at the vegetal pole of the egg.

Adult Strongylocentrotus purpuratus collected along the Southern California coast were maintained in chilled sea water tables so that gametes were available year round at the Caltech Kerckhoff Marine Laboratory (Leahy, 1986). Preparations of gametes, fertilization and embryo culture were carried out according to standard methods. The embryos were microinjected as described by McMahon et al. (1985) with gel-purified, linearized plasmid DNA. Approximately 2500 molecules of the desired plasmid DNA were introduced per embryo, together with a four-fold molar excess of PstI-digested carrier sea urchin DNA (Franks et al., 1990).

All of the expression constructs used in this work are illustrated diagrammatically in Figs 1 and 4, to which the reader is referred for nomenclature used in the following. Constructs Δ( −200), Δ( −155) and Δ( −10) were gel-purified after digestion of Δ( −440) DNA with HpaII, AvaII and DdeI, respectively. Δ( −440) is the same as pB2S•CAT of Sucov et al. (1988), which contains a BglII-SalI fragment of the original SM50 gene. Fusions Δ(d), Δ(b), Δ(a) and Δ(ab) were reconstructed from fragments of pB2S•CAT double digested with AvaII and DdeI; NsiI and SfuI; SfuI and SalI; and NsiI and SalI, respectively (see Fig. 3 for sequence; relevant restriction sites in this region of the SM50 gene are indicated in Fig. 1).

Mutations were introduced in the form of synthetic oligonucleotides. Essentially, the region to be mutated was in each case replaced by synthetic double-stranded oligonucleotides in which the sequence had been transversionally changed (i.e., A↔C; T↔G). Some oligonucleotides contained flanking restriction enzyme target sites also present in the SM50 gene so that they could be ligated into the expression vector. In other cases five-nucleotide tags were built onto the termini of the oligonucleotides, and complementary tags onto the termini of other oligonucleotides, for use in constructs bearing multiple mutated sites (cf. Fig. 1). The exact sequences of the SM50 regulatory domain that were transversionally mutated in this fashion are indicated in Fig. 3.

Other fusion genes utilized (Fig. 1D,E; and all constructs shown in Fig. 4) were assembled from vector elements and positive regulatory elements derived from different genes, together with oligonucleotides representing natural elements of the SM50 sequence. The sequences reproduced in these oligonucleotides are indicated as open boxes in Fig. 3. These oligonucleotides also contained terminal tags to allow their insertion into the vector or ligation to contiguous oligonucleotides. The sequence of the plus-strand oligonucleotide used for region D was cgcgCCAGGGTTACGacaccct; for region C, cgaggTGGTAGTCGTGAATGCATCGATCTCcgggtctag; and for region A, atcccAAAGTCTAGTGAGATCGCAACACATTTGA- GAAGCAg. The minus strand of the region A oligonucleotide, which is not shown here, contained a SalI site at its 5′ end. The elements to which these synthetic SM50 sequences were attached were obtained as follows. The SV40 early promoter was obtained from Promega, and was used as a BglII-HindIII fragment which has no enhancer (Gorman et al., 1982). Construct RTB-1 (Fig. 4) consisted of the CAT expression vector as above, the SV40 early promoter, and two fragments from the CyIIIa gene containing the distal SpGCF1 (P8) cluster of enhancer sites (Thézé et al., 1990; Zeller et al., 1995), and the sites at which the positively acting factors SpCTF1 (P4), SpTCF1 (P5), and SpOct1 (P3B) bind (reviewed by Coffman and Davidson, 1992). These three sites were obtained on a single restriction fragment from a deletion mutant of CyIIIa•CAT which lacks the negatively acting P3A2 target site (unpublished experiments). RTB-1 also includes a synthetic polylinker, PL1/2, containing ApaI, NheI, AscI, Acc65I, and BamHI sites, positioned at the 5′ side of the CAT coding sequence. A second synthetic polylinker, PL3/4, that includes HindIII, XhoI, ApaI, PstI, SphI, BglII, BsiWI, MluI, XbaI, EagI, EcoRV, and SalI sites, was positioned at the 5′ end of the SpGCF1 site cluster.

Synthetic TATA and initiator box elements were also used in some constructs. Based on consensus compilations (Kaufmann and Smale, 1994) the oligonucleotides bearing these included the sequences CATATTA, and ACACC, respectively. These sequences were built into a D′ or D oligonucleotide (cf. Fig. 3) and ligated together with other oligonucleotides and/or the vector, as required, by the same approach as above. In the synthetic TATA box constructs the junction point between the natural SM50 sequence and the TATA-D oligonucleotide was after the T at −55 of the natural sequence (Fig. 3).

The whole-mount in situ hybridization protocol used here is based on the method described by Ransick et al. (1993). However, the procedure was slightly modified by the use of Streck Fixative (Streck Laboratories Inc., NE). Fixation was carried out overnight at 4°C, followed by a brief wash in 50 mM Tris buffered-saline, containing 0.1% Tween 20. The proteinase K treatment and post-fixation procedure of Ransick et al. (1993) were omitted. CAT assays were carried out according to Seed and Sheen (1988).

Spatially accurate embryonic expression of exogenous SM50•CAT fusion constructs was demonstrated earlier by Sucov et al. (1988). Their starting construct, which was expressed exclusively in the skeletogenic mesenchyme cells of mesenchyme blastula-stage embryos, included SM50 sequences extending from −2200 downstream to +120, with respect to the transcription start site. Expression was assayed by radioactive in situ hybridizations carried out on serially sectioned samples. No change in level or locus of expression was observed when the upstream sequence was deleted down to −440, but further deletions sharply decreased expression (Sucov et al., 1988). This evidence provided the starting point for our present experiments. As we describe below, the minimal SM50 regulatory domain defined by Sucov et al. (1988) contains a number of functionally significant sequence elements, and their further analysis required the use of fine-scale deletions, mutations, and a variety of synthetic constructs. Fig. 1 displays a series of mutationally altered SM50•CAT constructs that we injected into zygotes and analyzed quantitatively for expression during embryogenesis. We had the advantage in this work of more sensitive methods of analysis than were available to Sucov et al. (1988), viz the whole-mount in situ hybridization procedure of Ransick et al. (1993; slightly modified, as described in Materials and Methods, by the use of a different fixative).

The construct shown at the top of Fig. 1, Δ( −440), serves as the control for the remaining constructs. In Fig. 1 the regulatory sequence included in Δ( −440) is divided into four color-coded subregions, to which we refer in the following. These are the ‘distal’ region (−440 to −200), shown in yellow; the ‘proximal’ region (−200 to −10), shown in green; the ‘initiator’ region (−10 to +10) shown in purple; and the ‘plus’ region (+10 to +120) shown in blue. There may well exist additional cis-regulatory sequence upstream of −440 in the native SM50 gene, particularly since we have not investigated requirements for larval or post-metamorphosis expression of this gene, which is utilized in adult test and spine-producing cells as well as in embryonic skeletogenic cells (Richardson et al., 1989). With respect to embryonic expression, however, we have confirmed and extended the observations of Sucov et al. (1988) that the SM50 sequence between −440 and +120 suffices to generate a normal pattern of expression. Thus, injected Δ( −440) is expressed almost exclusively in skeletogenic mesenchyme cells. Illustrative embryos in which skeletogenic mesenchyme expression of Δ( −440) is visualized by whole-mount in situ hybridization are shown in Fig. 2A1-A3.

When DNA is injected into sea urchin zygotes it rapidly concatenates in the cytoplasm, and the concatenate becomes stably incorporated in a blastomere nucleus, in most cases at 2nd, 3rd, or 4th cleavage, though later incorporations also occur (McMahon et al., 1985; Flytzanis et al., 1985; Hough-Evans et al., 1988; Livant et al., 1991). Thereafter the exogenous DNA replicates together with the host cell DNA (Franks et al., 1988; Livant et al., 1991), and is thus present in all of the cells of the clone to which the blastomere initially incorporating it gives rise. The transcriptional capacity of injected constructs is amplified by this process, because the exogenous DNA in each embryonic cell consists of a replica of the original concatenate. Thus as we found earlier (McMahon et al., 1985; Flytzanis et al., 1985; Livant et al., 1988), each cell bearing the exogenous DNA typically contains a sufficient number of copies of the introduced regulatory sequences to sequester a significant fraction of the transcription factors that service it. This means that the observed expression is contributed by many copies of the injected gene.

The mosaic incorporation pattern that results can be observed in Fig. 2A, for example. Eight of the 32 skeletogenic mesenchyme cells in the embryo at this stage are stained (these cannot all be distinguished clearly in the plane of focus shown), indicating that the incorporation occurred either in one of the first four blastomeres, in a vegetal 3rd-cleavage blastomere, or in a 4th-cleavage micromere, but not later. Embryos were considered positively stained only if two or more cells displayed the blue reaction product, and if the embryos were morphologically normal. In our experience the single stained cells that are occasionally observed usually indicate unstable or transient exogenous DNA that will disappear later in development, and that is often ectopically expressed irrespective of the construct injected. Single stained cells can also be produced as a result of experimental damage or abnormal developmental process that causes cells to delaminate from the ectoderm and fall into the blastocoel. The top line of data in Fig. 1 shows that by these criteria Δ( −440) was expressed in an average of close to 25% of embryos, based on examination of almost a thousand embryos belonging to 19 different batches of eggs (each batch is derived from a single female). Some batches were more active than others, and in these as much as 30-35% of embryos displayed labeled mesenchyme cells. The majority of batches expressed in about 24-28% of embryos. However, these values are all about two-fold lower than what would be expected on a purely random basis, judging from results obtained with constructs that are expressed in ectodermal or vegetal plate territories, in which 60-80% of embryos commonly express the injected constructs (C.-H. Yuh, C. Kirchhamer, E. Davidson, unpublished data). It follows that there is probably a bias against incorporation into skeletogenic precursors, i.e., the 4th or 5th cleavage micromeres. Such a bias might arise simply from the reduced chance these cells have of inheriting the exogenous DNA concatenate because of the disproportionately small fraction of zygote cytoplasm they inherit, or perhaps because of their delayed cleavage. An indication of a bias toward later incorporation than seen with other territories is that the average number of stained cells per positive embryo was, for Δ( −440), only three to four (again there was batch variation in this statistic). We note that Sucov et al. (1988) reported a similar result, about five cells labeled per positive embryo, based on a very much smaller sample.

Spatial expression of Δ( −440) is quite accurate, in that < 2% of injected embryos ever displayed any ectopic expression, i.e., expression other than in skeletogenic mesenchyme. Δ( −440) also contains cis-regulatory sites that mediate the temporal activation of the reporter as development proceeds. Transcripts of the endogenous SM50 gene accumulate sharply during development, from a few per skeletogenic precursor cell in late cleavage to 100-200 per cell from the mesenchyme blastula stage onward (Killian and Wilt, 1989). Measurements of the relative transcription rate by the run-off method also show that the SM50 gene is transcriptionally activated by mesenchyme blastula stage (Killian and Wilt, 1989), but no comparisons between mesenchyme blastula and later stages are available. Table 1 shows a several-fold increase in CAT enzyme content in embryos expressing Δ( −440) between mesenchyme blastula and 30 hour (early gastrula) or 48 hour (late gastrula) stages. The same fraction of embryos was observed to be expressing Δ( −440) at 21 hours, as at the 48 hour late gastrula stage for which data are recorded in Fig. 1. Therefore the number of expressing embryos at 48 hours is not limited by the sensitivity of the in situ hybridization (see also below). The developmental increase in CAT levels between mesenchyme blastula and gastrula stages requires a proportionately increased rate of transcription from Δ( −440) over this period, since both CAT mRNA and CAT enzyme protein are unstable in sea urchin embryos. Their half-lives were estimated earlier at ≤40 minutes (Flytzanis et al., 1987).

Preliminary experiments of Sucov (1989) showed that deletions extending proximally of −440 depress the level of expression of SM50•CAT constructs. Table 1 demonstrates that deletion of the distal region (construct Δ( −200) in Fig. 1) in fact results in a two- to three-fold decrease in CAT expression, as monitored at early and late gastrula stages. This, however, does not bring the level of CAT expression down near the limit of detection of whole-mount in situ hybridization, since, as shown at the right side of Fig. 1, about the same fraction of embryos bearing the Δ( −200) construct score as positive as of embryos bearing Δ( −440). Expression of Δ( −200) remains entirely confined to skeletogenic mesenchyme cells, as illustrated in the embryo shown in Fig. 2B1,2. Thus the only function that we observed for the distal region is quantitative. We did not explore further the distal region in this work, and thus it remains undetermined whether its positive function is specific to skeletogenic mesenchyme lineages. However, a clear conclusion from the Δ( −200) experiments can be drawn, viz that skeletogenic mesenchyme-specific regulatory elements are present in the SM50 sequence that lies between −200 and +120.

A further deletion to −155 gave essentially the same result as did Δ( −200), but when the sequences inward to −10 were also deleted, expression dropped precipitously (Fig. 1; Δ( −10)). Less than 10% of embryos now expressed CAT DNA at a level sufficient for detection by whole-mount in situ hybridization. Nonetheless the remaining expression was still confined to the skeletogenic mesenchyme. Therefore cis-regulatory elements specifying the territorial spatial function of the SM50 gene must exist still further downstream, between −10 and +120.

The positive regulatory function indicated by this experiment for the sequence between −155 and −10 (for convenience this region is labeled ‘d’ at the top of Fig. 1) cannot be substituted by the positive function resident in the distal region. Thus expression of construct Δ(d), which consists of Δ( −440) minus region d, is depressed as much as expression of Δ( −10) (see Fig. 1). To identify the responsible sequences we generated the twelve mutations labeled m(prox1) to m(prox1235) in Fig. 1. These were assembled from a series of overlapping synthetic oligonucleotides, as described in Materials and Methods. The exact sequences that were mutated by this means to produce the constructs shown in Fig. 1 are indicated in Fig. 3. As summarized in Fig. 1B, none of the single block mutations (i.e., m(prox1) to m(prox5)) affected expression as severely as observed with Δ( −10) or Δ(d), and mutants m(prox1) to m(prox3) were expressed as well as the control Δ( −440). Again in these experiments expression was always confined to cells of the skeletogenic mesenchyme lineages. Double, triple and quadruple block mutants were then tested. We found (Fig. 1B) that only in mutants in which the sequence of regions 4 and/or 5 were altered was the percentage of positive embryos significantly depressed (i.e., m(prox4), m(prox345), m(prox45), m(prox1234), and m(prox5), m(prox345), though not m(prox1235)). This result indicated that a strong positively acting element is present in both subregions 4 and 5. When both of these subregions are mutated (e.g. in m(prox4,5), expression is most severely affected, approaching the low level observed when the whole of region d is deleted. Another mutation, in which the sequence lying between subregions 4 and 5 was changed (m(prox123J); J, junction sequence) was expressed at the control level. It follows that subregions 4 and 5 may each include a target site for a positively acting factor, separated by the non-essential intervening J sequence (see Fig. 3). Subregions 4 and 5 apparently both contribute to the overall positive function of the proximal region. In context, subregion 4 might be the more essential, since m(prox1234) displays a depressed level of expression, while m(prox1235) expresses normally.

Fig. 3 shows that subregions 4 and 5 in fact both contain, on either side of the J sequence subelement, one occurrence of the short, directly repeated sequence AGGGTT. These sequences are labeled ‘D₍₄₎ repeat’ and ‘D₍₅₎ repeat’ in Fig. 3A. To test the proposition that the D repeat is responsible for the regulatory function of the proximal region revealed in these experiments, we constructed a synthetic oligonucleotide consisting of three copies of the sequence labeled D₍₄₎ in Fig. 3A (this includes two flanking nucleotides on the 5′ end of the D₍₄₎ repeat and three additional nucleotides on the 3′ end; these differ between D₍₄₎ and D₍₅₎). The synthetic sequence was then inserted into the Δ(d) construct (see Fig. 1, Construct 3xD), and injected into eggs. This construct behaved exactly as did the control Δ( −440), producing skeletogenic mesenchyme-specific expression in 27.4% of experimental embryos. An example is shown in Fig. 2C. The synthetic D₍₄₎ sequences thus account completely for the function of the proximal region of Δ( −440).

A similar analysis of the ‘plus’ region was carried out next. Results from the initial three constructs, shown diagrammatically in Fig. 1C as Δ(b), Δ(c) and Δ(ab), suggested that there are at least two subregions downstream of the transcription start site that are individually required for full expression, and that in their combined absence the construct is dead. We then built and tested the series of block mutations shown in Fig. 1C (see Fig. 3A for exact sequences), viz constructs m(plus1) to m(plus8). Mutations of a sequence extending from +10 to +30 entirely obliterated expression (construct m(plus1), while mutation of the adjacent seven nucleotides (construct m(plus2) depressed expression. In the following we designate the combined sequence that was mutated in these two constructs (i.e., +13 to +37) as region C. At least a major fraction of this sequence is absolutely required for SM50 fusion gene expression.

Proceeding downstream, the regions mutated in constructs m(plus3) and m(plus4), i.e., from +40 to +60 (Fig. 3), are apparently not essential. Thus as summarized in Fig. 1C, these constructs were expressed at normal levels, i.e., in about 24- 30% of injected embryos. However, mutations in the region in Fig. 1C for constructs m(plus5) to m(plus8). The sequence from +60 to +103 is designated region A in the following; like region D this sequence acts positively, but neither deletion nor mutation of region A affects the nearly perfect localization of the residual expression in skeletogenic mesenchyme cells. Normal expression of the construct shown at the bottom of Fig. 1C, m(plus/mid), confirms that regions C and A alone can substitute for the whole plus region, and also that the sequence between the C and A elements is dispensable.

By process of elimination it would appear from these experiments that the sequence element conferring skeletal mesenchyme-specific spatial expression must be either region C or the initiator region. All other deletions and mutations so far reviewed, i.e., the distal region deletions, the mutations that identified the D region of the proximal domain, and the A region mutations of the plus domain, preserve skeletogenic mesenchyme specificity. This test could of course not be applied to the mutation defining the C region, because without the C sequence no expression that can be observed by whole- mount in situ hybridization is obtained at all.

The SM50 promoter lacks a TATA box (Sucov et al., 1988), and instead belongs to the class of genes that utilize only an initiator sequence. This is the element that in a TATA-less promoter determines the transcription start site by serving as the assembly point for the TFIID complex (Kaufmann and Smale, 1994). To determine if the region in SM50 that extends from −10 to +10 with respect to the transcription start site might also be required for the spatial regulation of SM50•CAT constructs, we rebuilt this region of the gene, replacing it with a synthetic oligonucleotide encoding a consensus initiator site (Kaufmann and Smale, 1994). As shown in Fig. 3B, the natural SM50 sequence and the sequence including the initiator element (ACACC) differ completely for the 10 bp following the SM50 transcriptional initiation site (+1). A consensus TATA box sequence was also inserted immediately upstream of a synthetic D repeat element at position −25 (Fig. 3B). The SM50 sequence was thus changed into that of a typical TATA box-containing gene. The expression vector carrying this alteration, m(inr)TATA (Fig. 1D), was injected into eggs and its expression assessed by whole-mount in situ hybridization. Expression was again observed only in skeletogenic mes- enchyme, as illustrated in Fig. 2D. Furthermore, the efficiency with which m(Inr)TATA was expressed (21.0%) is only a few percent below that of the control Δ( −440). The original initiator sequence is therefore not the locus of spatial specificity in constructs such as Δ( −10) or m(plus/mid) (cf. Fig. 1).

To confirm that the D, C, and A elements function as well with the TATA-containing construct as with the natural initiation sequence, we carried out the experiments summarized in Fig. 4A. Here the entire sequence of Δ( −440) from −55 to +120, was replaced with synthetic DNA. As Fig. 3 indicates, −55 is the position at which the D repeats of the proximal region begin in the natural SM50 sequence. A 7 bp CATATTA sequence was inserted at this point, followed by the synthetic D sequence, and 25 bp from the TATA sequence, the initiator sequence (ACACC) was inserted, together with a restriction site linker to join it to the following C sequence, as above. The sequence of the junctional region of this vector, SM50•TATA, is shown in Fig. 3C. In it were mounted either normal, or transversionally mutated D, C, and A sequence elements in various combinations, as shown in Fig. 4A. When normal D, C, and A sequences were present, the construct functioned exactly as does Δ( −440), producing skeletogenic mesenchyme-specific expression in about 29% of experimental embryos. This result has two consequences. First, it proves that there is no detectable difference in transcriptional regulatory function whether the gene does or does not contain a TATA box sequence; second, it eliminates any possibility that sequence other than that included in the synthetic D, C, and A oligonucleotides is required for the regulatory function we observe. The further experiments, detailed in Fig. 4A, reproduce, with these entirely synthetic sequence elements, the same relationships seen in Fig. 1C. Thus, without a normal C sequence, expression is reduced to the level of the ectopic background expression, about 2% (this is shown by construct SM50•TATA(DA)). Addition of either the A or D oligonucleotide (SM50•TATA(DC) and SM50•TATA(AC)) to C partially restores expression, though D is more effective than A. Without either D or A, the C sequence element alone is not functional (SM50•TATA(C)). Thus it appears that for the C sequence element to promote expression, an ancillary positive function mediated by the D and/or A sequences is required. This is so even though the SM50•TATA constructs all include the distal region enhancer.

We next examined the function of synthetic D, C, and A elements in vectors that contained no other SM50 sequence. The construct SV40(DCA) (Fig. 1E) consists solely of the 200 bp enhancer-less SV40 early region promoter, which includes a TATA box and six Sp1 sites (Khoury and Gruss, 1983; Briggs et al., 1986), plus the synthetic D, C, and A oligonucleotides. While several transcription factors have been isolated from sea urchin embryo nuclear extracts that bind GC- rich target sites (Zeller et al., 1995; Hapgood and Patterson, 1994; Xiang et al., 1991) a sea urchin Sp1 factor has so far not been identified. Three of the Sp1 sites in the SV40 element, however, are also target sites for the sea urchin SpGCF1 factor (Zeller et al., 1995), which in context of various other factors functions as a positively acting regulator. Nonetheless, a vector consisting only of SV40 and the CAT reporter (SV40•CAT) has very little activity. The SV40(DCA) construct does express, however, and its expression is confined to skeletogenic mesenchyme. A quantitative comparison of CAT enzyme activity in embryos bearing SV40(DCA) and Δ( −440) is shown in Table, 2. While the level of SV40(DCA) expression is only 16% of that produced by Δ( −440), it is still sufficient to produce significant whole-mount in situ hybridization signals in about 18% of the embryos. An embryo bearing SV40(DCA) is shown in Fig. 2E1,2.

In the course of the foregoing experiments we never observed significant ectopic expression; that is, none of the deletions or mutations we studied provided even a hint of the existence of negative spatial regulatory functions in the SM50 cis-control system. The fact that SV40(DCA) is also expressed specifically in skeletogenic mesenchyme\ confirms that the spatial regulatory function carried out by the D, C, and A sequence elements is entirely positive. There are no other SM50 sequences present in this construct. Therefore there is no possibility that the accurate expression that it produces could be due to sequences outside of D, C and A that mediate negative control functions which we somehow missed in the foregoing experiments.

The control system isolated in SV40(DCA) also mediates the temporal activation of transcription seen earlier with Δ( −440). Thus as Table 1 shows, the CAT enzyme content of embryos expressing this construct is 5- to 7-fold higher at gastrula stage than at 21 hour mesenchyme blastula stage. Note that at all three stages there is about a six-fold ratio of expression of Δ( − 440) to SV40(DCA). This suggests that the change in expression that we observe over this developmental period is entirely due to the DCA element, and not to other regions of Δ( −440), while the absolute level of expression at each stage does depend on the positive elements located elsewhere in the regulatory domain. From the experiments reviewed above, these positive elements are likely to be the second D element that is present in Δ( −440), and the distal region element.

To compensate for the relatively weak expression of SV40(DCA) we added to it two DNA fragments bearing positively acting enhancers from the regulatory domain of the CyIIIa gene, which is normally expressed only in the aboral ectoderm of the embryo. These are, a distal region of CyIIIa bearing a cluster of the SpGCF1 sites, and a second fragment bearing sites for three different positively acting factors, SpOct1, SpCTF1, and SpTEF1 (Thézé et al., 1990; Calzone et al., 1988; Coffman and Davidson, 1992; C. Kirchhamer, J. Xian and E. Davidson, unpublished data). None of these regulatory factors are believed to convey any territorial spatial regulatory information or to be spatially confined in their activity. These positive regulatory elements were associated with SV40(DCA) with the use of an inserted polylinker (see Materials and Methods). The resulting construct, RTB-1(DCA), expressed almost as well as Δ( −440), as shown in Table 2. Expression was observed in 21% of experimental embryos (Fig. 4B), and was again specific to the skeletogenic mesenchyme. In the absence of the DCA insert, RTB-1 produces only a low level of (mainly ectodermal) expression in a few percent of embryos developing from injected eggs. When the synthetic DCA insert was reversed in orientation, and positioned at the distal end of the RTB-1 construct (see Fig. 4C), no activity above that of RTB-1 alone was observed.

Finally we sought to determine whether the D, C, or A element alone could confer spatial specificity, based on the theory that RTB-1 might include sufficient positive enhancer functions to display the spatial specificity of any added element if it had one. The experiments summarized in Fig. 4B provide the answer. RTB-1(C), in which the only non-mutated SM50 sequence is a single C element, is expressed specifically in skeletogenic mesenchyme, in 14.6% of embryos. Examples of such embryos are shown in Fig. 2F1-3. In the absence of element C no significant expression at all is observed (RTB- 1(DA), RTB-1(D), RTB-1(A); see Table 2, and Fig. 4B). Thus, the role of elements A and D in RTB-1(DCA) is merely to increase the activity of element C, and together with element C each of these is individually capable of increasing the level of expression (Table 2; constructs RTB-1(DC) and RTB- 1(CA)). These experiments demonstrate that element C alone is able to convey spatial specificity for transcriptional expression in the skeletal mesenchyme lineages.

The region of the SM50 regulatory domain that most of this study has focused on extends from −55 to about +105 with respect to the transcriptional start site. This sequence is capable of directing skeletogenic mesenchyme-specific expression of a reporter gene, at a level of activity that is only a few-fold below that of any of the longer constructs studied by Sucov et al. (1988) or ourselves. We show that the essential elements within this region are the directly repeated D sequences just upstream of the initiator element; the C element, which lies within the sequence from +12 to +37, and the A element, which lies within the region from +71 to +105. Thus, the C and A sequence elements are included in the transcribed leader sequence of the gene, and the translational initiation site occurs only a few bp beyond the end of the A regulatory element. We have established the functions of the D, C, and A elements by the use of synthetic oligonucleotides that represent either normal or mutated variants of the sequences. These were used to rebuild the −55 to +105 region of the expression constructs in a variety of combinations, in the context of both the normal upstream SM50 sequence and of an entirely heterologous expression vector that contained no SM50 sequences other than the synthetic D, C, and A elements. In addition, we altered the promoter of the SM50 gene from a TATA-less to a TATA-box conformation by insertion of canonical TATA and initiator sequences, and we showed that this transformation has no effect whatsoever on the functions mediated by the D, C, and A sequence elements.

The C element is the single irreplaceable regulatory sequence of the minimal, skeletogenic lineage-specific cis-control system extending from −155 to +105. It is the only one of the three SM50 elements which is alone capable of directing expression to skeletogenic mesenchyme cells, provided that the vector contains other elements that serve as enhancers. Entirely heterologous enhancers can perform this function, as shown by the experiments with the heterologous RTB-1 vector (Fig. 4B and Table 2). The RTB-1 enhancers cannot be skeletogenic mesenchyme-specific, since they derive from the CyIIIa gene, in which they promote expression in the aboral ectoderm. In the natural SM50 gene the A and D sequence elements provide the ancillary enhancer functions that the C element requires for activity; the A and D elements are dispensable only if other positively acting elements are provided (compare in Fig. 4, SM50- TATA(C) with RTB-1(C)). The measurements in Table 2, combined with the prior evidence obtained with specific mutations of D and/or A, show that both of these elements can work with C, and that each can function independently of the other. C is also the only element mutation which completely kills the ability of SM50 constructs to express. Although in the natural SM50 sequence the C element lies immediately adjacent to the initiator sequence this is not essential for its function. Thus, in the RTB-1 series of constructs the C element is separated from the initiator sequence by almost 100 bp of SV40 sequence, plus the length of the synthetic normal or mutated D sequence element, plus a linker added to provide convenient restriction sites, i.e., by about 130 bp. Furthermore, in theSM50- TATA and RTB-1 constructs the D element is downstream of the initiator site, and is present only singly, while in the normal gene there are two copies of D upstream of the site of transcriptional initiation. Similarly, the A element in these constructs is adjacent to C, except for a linker sequence, while in the normal alignment it is separated from C by about 30 nonessential base pairs. Thus, there is clearly some flexibility in the range of spatial relations among these elements that remain permissive for function, as well as in their positions with respect to the assembly point for the basal transcription apparatus.

The C element must bind a transcription factor that is present,or is functionally active only in skeletogenic mesenchyme lineages, or it could bind a generally present factor that in turn binds a lineage-specific co-factor. The C element in any case serves as the central regulatory information processor, by means of which the SM50 gene is instructed that it is in a skeletogenic precursor cell. It is not by itself a ‘tissue-specific enhancer’ since it is completely inactive in the absence of other enhancer elements. We cannot exclude the possibility that the proteins interacting with the D and A elements are also present or functionally active only in skeletogenic mesenchyme cells, but our experiments clearly show that they need not be. It is unlikely that the C element is a binding site for a repressor that in skeletogenic mesenchyme cells is replaced by an activator, while elsewhere in the embryo the repressor serves to keep the SM50 gene off. Were this the case, constructs such as RTB- 1(DA), in which the C element is mutated, would be expected to display enhanced ectopic expression; instead all constructs lacking a C element express very poorly, if at all.

It is curious that the C element can function with entirely heterologous enhancers as well as with its own natural retainers, A and D. This makes the idea that the binding of the C proteins serves as a focus for cooperative binding of the ancillary enhancer proteins less likely. An interesting speculation that satisfies our observations is that the lineage-specific protein that binds the C element (or its enhancer-specific co-factor) interacts directly with the basal transcription apparatus. Its function would be to enable or cause the basal transcription apparatus to respond to a range of enhancers, which may be present in several spatial domains of the embryo. That is, a corollary of this idea is that the basal transcription apparatus is not active in mediating upstream enhancer functions in the absence of a signal from the factor binding to the C sequence element (or from a cofactor bound in turn to the C factor). Thus the powerful CyIIIa elements that are included in the RTB-1 construct are almost inactive in the absence of the C element, since by itself this vector produces only very low levels of expression, none of which is in the skeletogenic mesenchyme. In short, we propose that the primary spatial specificity of the SM50 gene may be determined by a ‘locator’ element, which acts as an obligatory facilitator of the transcriptional enhancement functions of other DNA binding factors (see Fig. 5). This may not be an uncommon mechanism in regulatory systems that control spatial expression during development. For example, expression of the Drosophila Adh gene in the fat body depends on a cis-element located near the transcription start site, that like the SM50 C-element confers responsiveness to more distant enhancers, though in their absence it fails to promote transcription (Fischer and Maniatis, 1988).

Our knowledge of the functional significance of the sequence upstream of −55 is limited. There do not appear to exist any additional transcriptional control elements in the region beyond the two D sequence elements (i.e., from −55 to −200; this is shown by the results obtained with the prox series of mutations summarized in Fig. 1B). Of the distal region we know only that it includes positively acting elements which if present mediate several-fold greater activity. However, our preliminary data indicate that this region contains many closely packed target sites for DNA binding proteins, and the function of these interactions remains entirely unexplored. It is important to note that we have in no way excluded the possibility that the distal region could contain a second skeletogenic mesenchyme-specific positive control system.

To our knowledge SM50 differs from every other gene so far studied that is expressed in a territory-specific manner in early sea urchin embryos, in that its spatial control system appears to lack any negative regulatory elements that prevent expression in other territories. Such elements are easily detected in gene transfer experiments, since deletions, mutations, or in vivo competitions affecting interactions at these elements result in the often spectacular ectopic spatial expression of test constructs. For example, the CyIIIa gene (Hough-Evans et al., 1990; Wang et al., 1995; C. Kirchhamer and E. Davidson, unpublished data), the Spec1 gene (Gan et al., 1990), the Endo16 gene (C.-H. Yuh and E. Davidson, unpublished data), and the SM30 gene (T. Frudakis and F. Wilt, personal communication) are all controlled in part by negative spatial interactions. We saw no case, in this study, of a mutation or deletion that produces ectopic expression. Furthermore, we can state unequivocally that the spatial DCA control system functions only in a positive way. This is a particularly remarkable feature of the SM50 gene, in that its regulatory domain contains no less than five sites at which bind a regulatory factor, SpP3A2, that is almost certainly a negative regulator of the CyIIIa gene (Zeller et al., 1995; Hough-Evans et al., 1990). Interactions at the P3A target sites of the CyIIIa gene repress oral ectoderm expression of this gene (C. Kirchhamer and E. Davidson, unpublished data). In the SM50 gene these sites are at −354, −248, −128, −114, and −34. The double site at −114 and −128 has a particularly high affinity for this factor (Calzone et al., 1991). Yet, as Fig. 1 shows, every one of these sites was removed by one of the deletions or mutations tested, singly or in combination, and no ectopic expression was observed. Note, for example, that the m(prox123J) construct lacks both the strong double site and the site at −34, and yet its expression is quantitatively and spatially normal manner. The function of the P3A sites in the SM50 gene is another mystery that remains to be explored. It is still possible that the SM50 gene has a negative control system in the distal region, if both a negative functional element, and a positive functional element, the activity of which it modulates, are located together within this region. However, we can conclude that no negative spatial control system exists between −200 and +120, and yet this cis-regulatory sequence suffices to produce very accurate skeletogenic mesenchyme-specific expression.

The precursors of the skeletogenic mesenchyme are autonomously specified, as reviewed in the Introduction. This is significant in the present context, because an implication of an autonomous specification process is that the specific gene expressions that result could be mediated by localized maternal regulatory factors. The factor binding to element C has now been isolated by affinity chromatography (which affords the possibility of securing any co-factors as well), and it will soon be possible to test this proposition directly. If the factor binding to element C were indeed localized, or were present in active form only in the polar region of the egg inherited by the skeletogenic precursor cells, this would explain the absence of a negative control system in the SM50 gene. That is, the initial territorial specification would be spatially confined by cytoplasmic localization process(es), rather than by intercellular signaling across territorial boundaries in the cleavage-stage embryo. The latter argument would hold whether the localized factor is a C-element binding transcription factor, a co-factor, a system that covalently activates these factors, or a factor that promotes the transcription of a C-element binding transcription factor or co-factor.

Under various experimental situations other blastomeres than the precursors of the skeletogenic mesenchyme can be induced to give rise to mesenchyme cells that secrete skeleton and express skeletogenic genes such as SM50. Thus isolated macromeres, which normally give rise only to vegetal plate and some ectoderm, will differentiate as embryoids that contain some skeleton (HÖrstadius, 1939; Khaner and Wilt, 1991); animal pole mesomeres, if cultured for several days do the same (Khaner and Wilt, 1990; Henry et al., 1989); and so do various combinations of isolated mesomeres and other blastomeres (Khaner and Wilt, 1990). A well known phenomenon that is particularly germane is the effect of LiCl on isolated animal pole blastomeres (Livingston and Wilt, 1989; Von Ubisch, 1929). This agent induces differentiation in prospective ectoderm cells from the animal half of the embryo of both skeletogenic mesenchyme and gut, and for the effect to occur, exposure to LiCl must take place during cleavage, i.e., the period when blastomere specification is initially occurring. One interpretation (Davidson, 1989) is that transcription factors required for skeletogenic gene expression are present globally but must be activated in order to become functional. Normally these factors would be presented in active form only in the skeletogenic mesomeres and their descendants. All the experimental operations that induce ectopic skeletogenesis may affect intercell signaling in one way or another, and thus might result in ectopic activation of a factor such as that binding to the C element of SM50. However, it is possible that in none of these experimental situations is there a direct conversion to skeletogenic fate. Instead the only relevant conversion may be to vegetal plate fate, which in each of these Alternative versions of this model include the possibility that the LF actually recruits the appropriate TAFs to the TFIID complex; and/or that the LF interacts with the TFIID complex by means of an adaptor protein which is the factor that is lineage specific, rather than the DNA binding LF, as shown here.0020examples is indeed also known to be induced by the experimental circumstances imposed. The vegetal plate gives rise to the archenteron and to secondary mesenchyme. When there is a deficiency of the normal primary skeletogenic mesenchyme, cells of the secondary mesenchyme are able to transdifferentiate to skeletogenic mesenchyme. This could account for the appearance of skeletal elements and skeletogenic gene expression, including SM50 expression, in all of the experimental circumstances that we cite.

Conversion of secondary to skeletogenic mesenchyme fate at the gastrula stage is a remarkable phenomenon. It can be induced directly by experimental removal of skeletogenic mesenchyme cells (reviewed by Ettensohn and Ingersoll, 1992). A possible implication is that the SM50 gene must be under some sort of negative control in the secondary mesenchyme since it is derepressed in these cells when conversion takes place. On the surface this is a proposition that runs counter to our suspicion that the SM50 spatial control system lacks negative control elements. However, at this late period of embryogenesis the presentation of skeletogenic transcription factors almost certainly depends on their transcription de novo. Therefore it would be this process that is under negative control in the secondary mesenchyme.

In summary, we wish to stress the view that detailed functional knowledge of the cis-regulatory systems controlling spatial gene expression in the embryo will lead immediately to the generation of an invaluable set of new experimental probes. These can now be used to illuminate the mechanisms of classically defined embryological phenomena, in our case the initial autonomous specification, and the experimentally induced conditional activation, of skeletogenic cell lineages.

We are grateful to Dr Ellen Rothenberg for critical review of the manuscript. This research was supported by the ONR Cell Biology Program, Grant N00014-93-1-0379. K. W. M. was supported by the Japan Society for the Promotion of Science and by ONR.