Abstract
Previous studies have located some twenty distinct sites within the 2.3 kb 5′ regulatory domain of the sea urchin Cyllla cytoskeletal actin gene, where there occur in vitro high-specificity interactions with nuclear DNA-binding proteins of the embryo. This gene is activated in late cleavage, exclusively in cells of the aboral ectoderm cell lineages. In this study, we investigate the functional importance in vivo of these sites of DNA-protein interaction. Sea urchin eggs were coinjected with a fusion gene construct in which the bacterial chloramphenicol acetyltransferase (CAT) reporter gene is under the control of the entire CyIIIa regulatory domain, together with molar excesses of one of ten nonoverlapping competitor subfragments of this domain, each of which contains one or a few specific site(s) of interaction. The exogenous excess binding sites competitively titrate the available regulatory factors away from the respective sites associated with the CyIIIa · CAT reporter gene. This provides a method for detecting in vivo sites within the regulatory domain that are required for normal levels of expression, without disturbing the structure of the regulatory domain. We thus identify five nonoverlapping regions of the regulatory DNA that apparently function as binding sites for positively acting transcriptional regulatory factors. Competition with a subfragment bearing an octamer site results in embryonic lethality. We find that three other sites display no quantitative competitive interference with CyIIIa · CAT expression, though as shown in the accompanying paper, two of these sites are required for control of spatial expression. We conclude that the complex CyIIIa regulatory domain must assess the state of many distinct and individually necessary interactions in order to properly regulate CyIIIa transcriptional activity in development.
Introduction
The Cyllla cytoskeletal actin gene of the sea urchin Strongylocentrotus purpuratus is transcriptionally activated late in cleavage in the presumptive aboral ectoderm territory of the embryo and, throughout embryonic development, the gene continues to be transcribed exclusively in the aboral ectoderm (Shott et al. 1984; Cox et al. 1986; Lee, 1986; Hickey et al. 1987). A necessary and sufficient cw-regulatory domain extending approximately 2.3 kb upstream of the Cyllla transcription start site, which promotes correct temporal and spatial activation of a covalently associated CAT reporter gene, has been identified by gene transfer experiments (Flytzanis et al. 1987; Hough-Evans et al. 1987, 1988; reviewed by Davidson, 1989). The Cyllla cis-regulatory domain includes about 20 sites where there occur in vitro high-specificity DNA-protein interactions. These have been located in gel shift mapping experiments with specific oligonucleotide competitors, using crude embryonic nuclear extracts (Calzone et al. 1988; Thézé et al. 1990). This set of sites includes 13 wholly non-homologous, distinct recognition sequences, some of which occur several times in the Cyllla regulatory domain. There are unlikely to be additional binding sites in any of the fragments used in the competition experiments beyond those already characterized, that is, for factors that display a meaningful level of specific binding in vitro. Our objective in the present work was to obtain evidence in vivo of the functional significance of individual Cyllla proteinbinding sites, or of small subsets of locally contiguous sites. The approach chosen was based on the prior in vivo competition experiments of Livant et al. (1988). There it was shown that coinjection with the Cyllla-CATfusion of low (2 – 20 ×) molar excesses of the whole regulatory domain per se (i.e. lacking the CAT reporter gene) resulted in a near stoichiometric, quantitative decrease in reporter gene activity as a function of regulatory sequence: CAT reporter fusion gene ratio (molar excess, as used here, refers to the number of copies of competitor sites compared to CylIIa · CAT genes.) In the studies described in this and the accompanying paper, we coinjected restriction fragments containing subregions of the regulatory domain, together with a near saturating number of complete Cyllla · CAT genes. The significance of individual target sites (or small sets of target sites) on these subfragments was determined by assessing the activity of the CAT reporter gene later in development. From this approach two forms of information can be derived. When the coinjected restriction fragments contain sites of positive DNA-factor interaction that in vivo are required for normal gene function, competition for these specific factors depresses the level of CAT activity measured in the transgenic embryos. In the accompanying paper, we show that ectopic spatial expression results from competition with other specific regulatory target sequences, implying that the interactions occurring in these sequences function in vivo to control spatial expression.
Materials and methods
Preparation and injection ofS. purpuratus eggs
Unfertilized eggs were microinjected as previously described (McMahon et al. 1985). Briefly, the eggs were dejellied by exposure to pH5.0 filtered sea water (FSW) and fixed electrostatically to protamine sulfate-coated dishes. Approximately 2 pl of a DNA solution was injected into the cytoplasm of each egg. After injection the eggs were fertilized and cultured at 15°C, in the same dishes, in FSW containing penicillin (20 units ml-1) and streptomycin (50 μ gml-1).
CAT fusion genes and whole regulatory domain competitors
Two Cyllla · CAT fusion genes were used in these experiments. The 14kb Cyllla-CAT construct (Flytzanis et al. 1987) contains about 7 kb of genomic DNA extending 5′ from the transcription initiation site of the Strongylocentrotus purpuratus Cyllla cytoskeletal actin gene, plus the 5′ leader sequence of the Cyllla primary transcript that is interrupted by a 2.2 kb intron (Akhurst et al. 1987). The fusion point is a SalI site located 11 codons following the start codon of the Cyllla-CAT message. An SV40 poly (A) addition site and plasmid sequences are also included. The Cyllla-CAT construct was linearized for microinjection at a unique SphI site located about 2.5 kb upstream of the transcription start site (see Fig. 1). A second fusion gene was created by in-frame subcloning the 9.5kb SalI fragment from Cyllla-CAT (Fig. 1) into the CAT vector pUC.PL · CAT gene (Bond-Matthews and Davidson, 1988). The vector was first modified to create a SalI site (R. Franks et al. unpublished; Sucov et al. 1988). This fusion is called pCyllla · CAT · 1. The 6.3 kb SphI-KpnI fragment from Cyllla · CAT · 1 was gel purified and used for microinjection (Fig. 1). This fragment includes all necessary Cyllla regulatory sequences.
Structure of the pCyllla · CAT · 1 gene fusion, locations of sites of in vitro interaction between nuclear factors and Cyllla 5′-flanking sequence, and competitor DNA fragments. (A) Map of pCyllla · CAT · 1. Hatched areas represent sea urchin sequences, including the 2.2 kb first intron contained within the Cyllla primary transcript (Akhurst et al. 1987). The fusion point is a SalI site located 11 codons following the start codon of pCyIIIa · CAT · 1. An SV40 poly(A) addition site is included, together with plasmid sequences. The 6.3kb SphI-KpnI fragment derived from the pCyllla · CAT · 1 fusion gene was used for microinjection. Alternatively, the CyIIIa · CAT fusion gene (not shown) described by Flytzanis et al. (1987) was linearized at the SphI site located 2.5 kb upstream of the CyIIIa transcription start site, and used for microinjection. (B) Map of the CyIIIa cw-regulatory domain, and competitor DNA fragments. The locations of all known sites of in vitro interaction between sequence elements in the 5′-flanking regulatory region of the Cyllla gene and nuclear factors are depicted by black boxes within the 2.5 kb Pstl-BamHI fragment. The protein-DNA interactions described by Thézé et al. (1990) for each binding site are indicated (P1 – P8). The Pl binding region actually extends into the 5′ end of the first exon of the CyIIIa gene. The 9.5 kb CyIIIa Sad fragment and the 4.3 kb NcoI-EcoR1 fragment (indicated above the map), each of which contains the entire regulatory domain, were used as competitor in the in vivo competition experiments shown in Fig. 2. DNA subfragments of the regulatory region (shown below the map) that contain the indicated factor-binding site(s) (or no binding site, x) were released from pUC subclones (Calzone et al. 1988) by digestion from the polylinker with Eco R1 and HindlII (P1 – P3, P4 – P6, x) or HindIII (P5), or from pBluescript subclones (A. Cutting, E. Davidson, unpublished data) by digestion with BamHI and HzndIII (P3A, P7I). Fragment P3B was derived from fragment P3 by digestion with HinfI. These fragments were ligated to sea urchin carrier DNA molecules bearing termini homologous with the specific fragment and prepared for microinjection as described in Materials and methods, and were subsequently used for the in vivo competition experiments shown in Fig. 3 and Table 1. Restriction sites shown are: a, Avail; b, BamHl; d, Ddel; dr, Dral; h, Hwidlll; ha, Haelll; hi, Hind; hp, Hpall; k, Kpnl; n, Ncol; p, PstI; r, Rsal; e, EcoRI; s, Sall; sp, SphI.
Structure of the pCyllla · CAT · 1 gene fusion, locations of sites of in vitro interaction between nuclear factors and Cyllla 5′-flanking sequence, and competitor DNA fragments. (A) Map of pCyllla · CAT · 1. Hatched areas represent sea urchin sequences, including the 2.2 kb first intron contained within the Cyllla primary transcript (Akhurst et al. 1987). The fusion point is a SalI site located 11 codons following the start codon of pCyIIIa · CAT · 1. An SV40 poly(A) addition site is included, together with plasmid sequences. The 6.3kb SphI-KpnI fragment derived from the pCyllla · CAT · 1 fusion gene was used for microinjection. Alternatively, the CyIIIa · CAT fusion gene (not shown) described by Flytzanis et al. (1987) was linearized at the SphI site located 2.5 kb upstream of the CyIIIa transcription start site, and used for microinjection. (B) Map of the CyIIIa cw-regulatory domain, and competitor DNA fragments. The locations of all known sites of in vitro interaction between sequence elements in the 5′-flanking regulatory region of the Cyllla gene and nuclear factors are depicted by black boxes within the 2.5 kb Pstl-BamHI fragment. The protein-DNA interactions described by Thézé et al. (1990) for each binding site are indicated (P1 – P8). The Pl binding region actually extends into the 5′ end of the first exon of the CyIIIa gene. The 9.5 kb CyIIIa Sad fragment and the 4.3 kb NcoI-EcoR1 fragment (indicated above the map), each of which contains the entire regulatory domain, were used as competitor in the in vivo competition experiments shown in Fig. 2. DNA subfragments of the regulatory region (shown below the map) that contain the indicated factor-binding site(s) (or no binding site, x) were released from pUC subclones (Calzone et al. 1988) by digestion from the polylinker with Eco R1 and HindlII (P1 – P3, P4 – P6, x) or HindIII (P5), or from pBluescript subclones (A. Cutting, E. Davidson, unpublished data) by digestion with BamHI and HzndIII (P3A, P7I). Fragment P3B was derived from fragment P3 by digestion with HinfI. These fragments were ligated to sea urchin carrier DNA molecules bearing termini homologous with the specific fragment and prepared for microinjection as described in Materials and methods, and were subsequently used for the in vivo competition experiments shown in Fig. 3 and Table 1. Restriction sites shown are: a, Avail; b, BamHl; d, Ddel; dr, Dral; h, Hwidlll; ha, Haelll; hi, Hind; hp, Hpall; k, Kpnl; n, Ncol; p, PstI; r, Rsal; e, EcoRI; s, Sall; sp, SphI.
Competition in vivo between the Cyllla 5′-regulatory domain and coinjected Cyllla · CAT reporter genes. At least 1200 molecules of a Cyllla · CAT gene fusion were injected at each point, together with varying numbers of either the 9.5 kb SalI DNA fragment or the 4.3 kb NcoI-EcoRI DNA fragment, containing all known sites of in vitro interaction between the Cyllla 5′-regulatory region and nuclear factors (see Fig. 1B) as competitor. The molar ratio of competitor sequence to Cyllla · CAT sequence incorporated at each point was measured by slot-blot hybridization of DNA extracted from samples of pooled transgenic embryos, using 32P-labeled probes specific for each sequence, as described in Materials and methods. A second aliquot of embryos was pooled and assayed for CAT activity at each point. The relationship between CAT activity per embryo relative to the CAT activity per embryo (normalized to 1.0) in the control in which no competitor DNA was included (y), and molar ratio, (x), of competitor DNA molecules to Cyllla · CAT genes, can be approximated as y=(1+ α x)-1, where α is the competition efficiency coefficient, and provides an estimate of the fraction of competitor fragments actively functioning as competitor (Livant et al. 1988; see legend to Table 1). The dashed curve indicates the ideal form, i.e. where α =l.0. The solid curve is the least squares fit to the experimental data, and yields a value for a of 0.76 [Livant et al. (1988) obtained α =0.44 in similar experiments; considering the scatter there may be no real difference between these results, nor any real discrepancy from an α value of unity]. Twenty-six experiments are shown. Each unique symbol represents an experiment carried out with a single batch of eggs. Cyllla · CAT gene fusion and competitor DNA fragment used are: solid symbols (black or white), 6.3kb SphI-KpnI pCyllla · CAT · 1 fragment plus 9.5 kb SalI competitor DNA fragment; cross-hatched symbols, SphI-linear Cyllla · CAT (Flytzanis et al. 1987) plus 9.5 kb SalI competitor DNA fragment; right-shaded symbols, SphI-linear Cyllla · CAT plus 4.3 kb Ncol-EcoRi competitor DNA fragment.
Competition in vivo between the Cyllla 5′-regulatory domain and coinjected Cyllla · CAT reporter genes. At least 1200 molecules of a Cyllla · CAT gene fusion were injected at each point, together with varying numbers of either the 9.5 kb SalI DNA fragment or the 4.3 kb NcoI-EcoRI DNA fragment, containing all known sites of in vitro interaction between the Cyllla 5′-regulatory region and nuclear factors (see Fig. 1B) as competitor. The molar ratio of competitor sequence to Cyllla · CAT sequence incorporated at each point was measured by slot-blot hybridization of DNA extracted from samples of pooled transgenic embryos, using 32P-labeled probes specific for each sequence, as described in Materials and methods. A second aliquot of embryos was pooled and assayed for CAT activity at each point. The relationship between CAT activity per embryo relative to the CAT activity per embryo (normalized to 1.0) in the control in which no competitor DNA was included (y), and molar ratio, (x), of competitor DNA molecules to Cyllla · CAT genes, can be approximated as y=(1+ α x)-1, where α is the competition efficiency coefficient, and provides an estimate of the fraction of competitor fragments actively functioning as competitor (Livant et al. 1988; see legend to Table 1). The dashed curve indicates the ideal form, i.e. where α =l.0. The solid curve is the least squares fit to the experimental data, and yields a value for a of 0.76 [Livant et al. (1988) obtained α =0.44 in similar experiments; considering the scatter there may be no real difference between these results, nor any real discrepancy from an α value of unity]. Twenty-six experiments are shown. Each unique symbol represents an experiment carried out with a single batch of eggs. Cyllla · CAT gene fusion and competitor DNA fragment used are: solid symbols (black or white), 6.3kb SphI-KpnI pCyllla · CAT · 1 fragment plus 9.5 kb SalI competitor DNA fragment; cross-hatched symbols, SphI-linear Cyllla · CAT (Flytzanis et al. 1987) plus 9.5 kb SalI competitor DNA fragment; right-shaded symbols, SphI-linear Cyllla · CAT plus 4.3 kb Ncol-EcoRi competitor DNA fragment.
Subfragments of the Cyllla regulatory domain used as competitors
Subfragments that were used for the in vivo competition experiments were derived either directly from the Cyllla · CAT construct, or from pUC 18/19 subclones of the Cyllla regulatory region (Calzone et al. 1988; F. Calzone, R. Hill, E. Davidson, unpublished), or equivalent Bluescript subclones (A. Cutting and E. Davidson, unpublished) by digestion with the appropriate restriction enzymes (see legend to Fig. 3). The fragments were gel purified. Subfragments less than approximately 1.0 kb in length were ligated in vitro to sperm DNA (prepared as previously described; Lee et al. 1984). The sperm DNA was digested with an appropriate restriction enzyme to produce molecules of an average length of about 5.0kb, the termini of which would be homologous with the competitor DNA termini, or blunt-ended if the competitor DNA molecule bore a blunt end. Ligation reactions were performed at 15 °C overnight in 20 – 50 μ l that contained up to 10 μ g of total DNA with an average molar ratio of competitor DNA fragments/carrier DNA fragments of about 5.0. The ligated DNA was digested with BglII, an enzyme for which none of the competitor DNA subfragments included a site, and molecules ranging in size from approximately 1 kb to 15 kb were obtained for microinjection by fractionation in a 1% low-melt agarose gel, followed by purification by the Geneclean (Bio 101 Inc.) procedure. Alternatively, the BglII- digested DNA was size-fractionated by gel filtration through Bio-Gel A-5m (BioRad). The actual concentration of the specific competitor DNA sequence in the final sample used for injection was measured by slot-blot hybridization, using a single-stranded RNA probe. Known numbers of linearized Cyllla - CAT DNA molecules were included on each filter as standards. Typically, the competitor DNA sequence was present at an average density of about 3.4 copies per 5 kb of carrier DNA in the sample (all samples used fell within the range 2 – 6 copies per 5 kb carrier DNA except the P7I ligation, which was injected at 1 copy per 5kb). A few deviations are noted in the legend to Fig. 3.
In vivo competition by subelements of the Cyllla 5′-regulatory region in transgenic embryos. At least 1200 molecules of a Cyllla · CAT gene fusion were injected at each point, together with varying numbers of competitor fragments that contain the binding regions for the indicated set of DNA-binding factors (Thézé et al. 1990; see Fig. IB). The relationship between average CAT activity per embryo and the average molar ratio of competitor DNA molecules to Cyllla · CAT molecules incorporated is expressed as in Fig. 2. The competition coefficient α (i.e. where ideal stoichiometric competition would generate a value for α of 1.0) calculated for the whole regulatory domain in Fig. 2 was used as a normalization factor to correct the individual data points in these experiments (see legend to Table 1), so that the competitive effect of the individual fragments would be compared directly to the effect of the whole domain. Thus, the dashed curves in A–J indicate the competitive behavior of the entire regulatory region. The solid curves, fit by least squares, indicate the behavior of the individual subfragments. Competition is near stoichiometric in A–C. There is little, if any, competition in F–J. Panels A–C, E, G, H, J show the least squares fit to the pooled experimental data obtained from more than one experiment. The data for the eighteen experiments shown here were derived from a subset of those experiments used to generate the value of a for the whole regulatory domain shown in Fig. 2 (data points from a single experiment are described by matching symbols in these figures).
In vivo competition by subelements of the Cyllla 5′-regulatory region in transgenic embryos. At least 1200 molecules of a Cyllla · CAT gene fusion were injected at each point, together with varying numbers of competitor fragments that contain the binding regions for the indicated set of DNA-binding factors (Thézé et al. 1990; see Fig. IB). The relationship between average CAT activity per embryo and the average molar ratio of competitor DNA molecules to Cyllla · CAT molecules incorporated is expressed as in Fig. 2. The competition coefficient α (i.e. where ideal stoichiometric competition would generate a value for α of 1.0) calculated for the whole regulatory domain in Fig. 2 was used as a normalization factor to correct the individual data points in these experiments (see legend to Table 1), so that the competitive effect of the individual fragments would be compared directly to the effect of the whole domain. Thus, the dashed curves in A–J indicate the competitive behavior of the entire regulatory region. The solid curves, fit by least squares, indicate the behavior of the individual subfragments. Competition is near stoichiometric in A–C. There is little, if any, competition in F–J. Panels A–C, E, G, H, J show the least squares fit to the pooled experimental data obtained from more than one experiment. The data for the eighteen experiments shown here were derived from a subset of those experiments used to generate the value of a for the whole regulatory domain shown in Fig. 2 (data points from a single experiment are described by matching symbols in these figures).
Microinjection solutions
Each DNA solution was made up in 40% glycerol and contained 550 – 2000 molecules of a CAT fusion gene per pl, together with competitor DNA and/or carrier DNA (see text). The DNA mass in each injection solution was adjusted by the addition of carrier to yield at least a 5-fold mass excess of carrier DNA plus competitor DNA, over CAT DNA, and to ensure that each injection sample in a given experiment contained the same total DNA concentration.
Measurement of CAT enzyme content per embryo
Embryos were analyzed for CAT enzyme content at 20 – 24 h postfertilization (hatched blastula) as previously described (McMahon et al. 1984), using lysates derived from 40 – 60 injected embryos and a series of bacterial CAT enzyme standards. Under our conditions, one unit of CAT enzyme activity equals ∼ 2.6 × 1011 molecules of enzyme, as established earlier by reference to a purified recombinant CAT enzyme preparation (McMahon et al. 1984).
Measurement of CAT DNA and competitor DNA content per embryo
DNA measurements were made at 24 h postfertilization by a modification of the slot-blot procedure described earlier (Flytzanis et al. 1987; Franks et al. 1988). Centrifugal pellets containing 60 – 130 injected embryos or uninjected control embryos were mixed in 50 μ l of 250 mM Tris-HCl, pH 7.8 with ∼ 100 gastrula stage Lytechinus variegatus embryos used as carrier. An equal volume of 0.1 M EDTA, pH 8.0, 0.2 M NaCl, 1% SDS was added, and the samples were incubated with 20 μ g of proteinase K for 3h at 55 °C. The nucleic acids were extracted and treated with 0.4 M NaOH at 65°C to denature the DNA and hydrolyze the RNA, and were filtered onto nitrocellulose using a slot-blot apparatus as previously described. For hybridization standards known numbers of Cyllla-CAT DNA molecules, linearized with SphI, were mixed with L. variegatus carrier nucleic acids, treated with 0.4M NaOH as described above, and applied to each filter. These standards of course include all of the different competitive sequences, since the whole regulatory domain is present in Cyllla · CAT (see Fig. 1) as well as the CATgene sequence. The slots were cut exactly in half, and the halves were separately hybridized with single-stranded RNA probes containing sequences representing either the CAT gene (Flytzanis et al. 1987) or the Cyllla 5′-sequence used for in vivo competition. Probe specific activities were about 1.3 × 109ctsmin-1 μ g-1. Hybridizations with the CAT probe were carried out at 40 °C in 50% formamide as described (Franks et al. 1988). Hybridizations with the Cyllla 5′-competitor sequence probes were performed at temperatures calculated to maximize hybridization efficiency based on G-C content (Casey and Davidson, 1977). The filters were washed twice with 2 × SSC (1 × SSC is 0.15M NaCl, 0.015M sodium citrate), 0.2% SDS at room temperature, twice with 2 × SSC, 0.2% SDS at 68°C, and twice with 0.75 × SSC, 0.2% SDS at 68 °C. Following autoradiography, each half-slot was cut out and counted; quantitation of CAT DNA sequences or Cyllla 5′-competitor DNA sequences in each half-slot was made by comparison to the signal obtained for hybridization between each probe and the Cyllla · CAT DNA standards on each filter. The competitor DNA sequence probes also hybridized with the incorporated Cyllla · CAT sequences present in each half-slot of DNA extracted from the injected embryos. Therefore, the number of competitor DNA molecules in excess of Cyllla · CAT genes in a given half-slot could be determined by subtracting the number of Cyllla · CAT DNA molecules that was calculated to be present (using the corresponding half-slot hybridization signal obtained with the CAT probe) from the total number of sequences that hybridized with the competitor DNA probe. Thus the molar ratio of incorporated competitor DNA:CyIIIa · CAT DNA in the transgenic embryos was obtained. The number of CAT DNA molecules and competitor DNA molecules present per halfslot was also computed in terms of average numbers of molecules per embryo.
Competitive efficiency coefficient α
The competitive in vivo titrations carried out with the whole regulatory domain by Livant et al. (1988) indicated that the efficiency of competition in vivo was about 44% of ideal expectation. This was established by deriving a competition efficiency coefficient, α, given the form y=(l + α x)-1, where y is the CAT activity relative to the control, in which Cyllla · CAT but no competitor DNA is introduced into the embryos; and x is the molar ratio of competitor to Cyllla · CAT DNA actually incorporated in the experimental embryos. Were α =l, the factor binding sites included in the incorporated Cyllla-CAT genes, and those in the incorporated competitor regulatory domains, would have competed with equal efficiency for the limiting DNA binding factors. The microinjection gene transfer system utilized in both the experiments of Livant et al. (1988) and those described in this paper, produces a mosaic, though random, pattern of exogenous DNA incorporation (McMahon et al. 1985; Flytzanis et al. 1985; Hough-Evans et al. 1987; Franks et al. 1988; see Livant et al. 1988). Thus, a probable explanation for the ∼ 2-fold discrepancy from ideal behavior in the experiments of Livant et al. (1988) is that the cells incorporating and expressing Cyllla · CAT in their genomes are not invariably the same cells that incorporate all the competitor sequences.
Results
Competitive titration using the entire Cyllla regulatory domain
Previously published gene transfer studies demonstrating that correct spatial and temporal expression is mediated by upstream Cyllla sequence were carried out with the linearized Cyllla · CAT fusion (Flytzanis et al. 1987; Hough-Evans et al. 1987, 1988), and this construct was also used in some of the experiments we report here. A second construct, pCylHa · CAT · 1, utilized in other experiments, is shown diagrammati-cally in Fig. 1A. Expression of the 6.3 kb SphI-KpnI fragment of pCyllla · CAT · 1 in transgenic embryos is indistinguishable from that of Cyllla · CAT in quantitative, temporal, and spatial respects. On microinjection into sea urchin egg cytoplasm, linear DNA molecules of these lengths are rapidly ligated, irrespective of the DNA termini, into long end-to-end concatenates that are stably incorporated into one or more blastomere nuclei during cleavage, and are replicated together with host cell DNA throughout embryogenesis (McMahon et al. 1985; Flytzanis et al. 1985). No differences were detected in the degree of amplification of any of the fusion genes or competition fragments within experiments.
In an initial series of experiments, fragments containing the complete set of detectable protein-binding sequence elements were used as in vivo competitors in order to provide base line stoichiometry for comparison with the competitive effects of individual subfragments. In Fig. 2 we present new data that substantiate the observation (Livant et al. 1988) that quantitative competitive titrations can be obtained in vivo in sea urchin embryos. This Fig. shows the pooled ‘whole regulatory domain’ controls for the subfragment competition experiments presented in the following. For each point in Fig. 2, at least 1200 molecules of the Cyllla · CAT reporter, plus the indicated numbers of competitive DNA molecules (9.5 kb SalI or 4.3 kb NcoI – EcoRI fragments, see Fig. 1) were injected. At least a 5-fold mass excess of carrier and/or competitor DNA was coinjected with the Cyllla-CAT genes. The carrier used was sea urchin sperm DNA, cut to a mean length of 5 kb. Thus the individual Cyllla · CAT genes in the incorporated concatenates were distributed around a mean distance apart of 30 – 70 kb (depending on whether the 14kb Cyllla-CAT or the 6.3kb SphI-KpnI fragment was used), interspersed among randomly selected sequences; or, in the competition samples, among various ratios of Cyllla regulatory domain and 5 kb randomly selected sequence. Preliminary experiments had shown that the Cyllla · CAT gene functions 2 – 3 × more efficiently when separated by ⩾ 3.5-fold excess of carrier (possibly due to transcriptional interference in the absence of carrier). Within each experimental series of Fig. 2, the various competition samples and the reference sample, in which Cyllla · CAT and carrier but no competitor had been introduced, all received exactly the same total mass of DNA per egg; i.e. the carrier DNA content was adjusted to take into account the mass of competitor DNA added to obtain the nominal competitor/fusion gene molar ratio desired. The total DNA mass injected was about 0.2pg/egg.
As in our previous studies (op cit) average CAT enzyme content per embryo in the reference samples that contained no competitor was generally 1.0 – 3.0 × 107 enzyme molecules per embryo. To pool the data shown in Fig. 2 these values have been used for normalization of the CAT enzyme content in the embryos within each experimental series. The CAT enzyme was measured in samples of 24 h embryos, and in Fig. 2 is expressed as a function of the molar ratio of competitor to Cyllla · CAT genes actually incorporated per average embryo. These ratios were directly measured in aliquots of the same embryo samples by slot-blot hybridization, using competitor DNA and CAT DNA sequence probes. The experimental data fit by the solid line in Fig. 2 indicate that CAT activity falls off at a rate that is almost stoichiometric, with respect to the number of competing 5′ regulatory sequences present. The best fit value from the pooled data of Fig. 2 yields β =0.76 (see legend of Fig. 2). Thus we have used β =0.76 as a normalization factor in most of the following experiments. These data clearly confirm that endogenous factors required for the expression of the Cyllla · CAT fusion gene can be titrated away from the fusion gene by introduction of excess DNA binding sites.
Competitive titration in vivo by subelements of the Cyllla regulatory domain
To determine the effects on Cyllla · CAT gene activity of titrating away those factors binding to individual subregions of the large regulatory domain, we coinjected various molar excesses of the subfragments identified in Fig. IB. Many of these subfragments were also utilized for the gel-shift experiments of Calzone et al. (1988) by which the sites of interaction were initially located, and we have adopted the same subfragment nomenclature in the present account. Several include only a single site of DNA-protein interaction. However, the Pl, P2 and P6 subfragments contain two sites of interaction, and the P8 subfragment, which is particularly complex, includes at least 8 protein-binding elements of three different classes (Thézé et al. 1990). Two subfragments that do not contain sequences that form DNA-protein complexes in vitro on reaction with embryo nuclear extracts are also shown on the map (X in Fig. IB). Neither of these subfragments displayed any competitive function in our hands (results not shown), nor do subfragments extending 5.3 kb upstream of the Sphi site (see Fig. 1A). In initial studies we found that microinjection of small DNA fragments (<1kb) does not reliably result in measurable replication during embryogenesis, probably due to failure to form concatenates that are sufficiently large to promote stable incorporation. To circumvent this problem each of the competitor DNA subfragments shown in Fig. IB, with the exception of the large subfragment containing the whole P8 binding region, was first ligated in vitro to carrier genomic DNA. The actual concentration of the specific competitor sequence in each ligated DNA preparation was measured by slot-blot hybridization, using a 32P-labeled RNA probe.
In each injection series, eggs obtained from a single female were coinjected with at least 1200 molecules of a Cyllla · CAT gene, together with increasing quantities of competing sequence that were always in molar excess with respect to the Cyllla · CAT DNA. Competition curves obtained in vivo for each of nine nonoverlapping subfragments that were derived from the Cyllla regulatory domain are shown in Fig. 3. The average CAT enzyme content and the average competitor DNA:Cyllla · CAT DNA ratio were measured in pooled embryo samples at 24 h. The average exogenous DNA amplification in the various injection samples within a given experiment was similar, and in different batches the values measured in the 24 h embryos ranged from about a 10-fold to an 80-fold increase over the amount of DNA injected per egg. Normalized experimental data were fitted to the same form describing the competition stoichiometry as above, and for each experiment a value for the competition efficiency coefficient α was derived by nonlinear least squares. In Fig. 3, competition functions are shown for each of 10 subfragments that contain known sites of DNA-protein interaction identified in vitro. These are presented in order of descending avalúes; i.e. the best competitors first, and the least effective last. The curve representing perfect competition stoichiometry (α =1) is shown as a dotted line in each panel for comparison.,Fig. 3A – F demonstrates immediately that a majority of the nonoverlapping subfragments tested function individually as competitors against the whole Cyllla regulatory domain, and the P4, P5 and P8 subfragments each appears to compete as well as does the whole regulatory domain. Each of these regions apparently contains an essential positive regulatory site(s) bound by a factor or factors of limited availability relative to the number of competitor fragments incorporated. Fig. 3G-J shows that subfragments containing the binding regions for P6, P3, P3A and P7II, respectively, displayed relatively insignificant competitive activity.
Quantitative data for these and some additional experiments are fisted in Table 1. The values given by the parameter α ′ in the second column of Table 1 estimate the competitive effect of the indicated subfragment in each experiment, relative to the competition obtained with the complete regulatory region. Pooled data for multiple experiments with given subfragments were used to assess the best overall value of the competition coefficient, given as α ″ in the third column of Table 1. Each of five different regions, i.e. those containing binding sites for P2, P4, P5, P7I and P8, compete between 10 % and 100 % (or better) than does the whole regulatory domain (i.e. α′5=0.1). An α′ value of 1.0 would indicate that the competitive effect of the subfragment is equal to that of the fragment containing the entire regulatory domain. The α′ values for DNA subfragments containing the binding sites for P3, P3A, P6 or P7II are below 0.1, and that for Pl is just above 0.1. As discussed below, this result is unlikely to be due simply to high prevalence of the respective factors.
Lethal effects of competitive titration of the P3B binding site
The site labeled P3B in Fig. 1 includes a canonical octamer binding site (Thézé et al. 1990). Uniquely among all the subfragments tested, competition with a subfragment containing only this site results in the death of 60 % to over 95 % of the recipient embryos, by the pluteus stage, in different experiments. In controls for these particular experiments, 95–100% of eggs injected only with a saturating number of Cyllla · CAT molecules plus carrier DNA developed through hatching blastula stage, and 70–80 % completed embryogenesis normally. The same P3B injection solution that is lethal to recipient embryos is rendered harmless if treated with DNase prior to injection.
The unusual lethality of P3B competition is particularly interesting because as shown earlier the adjacent P3A site, when injected on a fragment by itself, does not display competitive activity (Fig. 31; Table 1), and when the Hináill-Ddel ‘P3’ subfragment, which includes both the P3A and P3B sites (see Fig. 1A) is injected, the results are the same as when the P3A subfragment alone is injected (see Table 1). That is, there is no significant competitive activity (α′ is about 0.04 for both subfragments), and no lethality is observed. The P3B factor is evidently required for the expression of genes that are necessary for viability, a not unlikely possibility since octamer binding sites have been reported in a great many different genes in mammals, and in several sea urchin histone genes as well (Barberis et al. 1987). That the presence of the P3A site on the same competitor fragment with the P3B site ‘rescues’ the embryos from the lethal effects of P3B competition suggests that, when P3A is bound at the adjacent sites, the essential P3B octamer factor is precluded from binding, and thus is not sequestered away on the competitor fragments.
Discussion
Competitive titration of positively acting factors within the embryo nuclei
Livant et al. (1988) showed that as the number of Cyllla · CAT genes per embryo nucleus is increased, over a range from a few to several thousand, the amount of CAT enzyme expressed attains a plateau, after an initial phase in which CAT activity rises proportionally with incorporated gene number. The kinetics of the response can be fitted to the expected form for a saturation phenomenon. This is an interesting result which it is necessary to consider briefly before taking up the nature of the competitions that we present in this paper. About 90 % saturation of Cyllla · CAT expression occurs at 1050 specific sites per nucleus, implying that the limiting factors required for Cyllla · CAT expression are 90 % bound to specific sites at this level (D. Livant and E. Davidson, unpublished). In the present experiments, we tried to approach saturation conditions by injecting ⩾ 1100 molecules of Cyllla-CAT per egg, plus in the competition samples, molar excess of the individual sites being tested. On the saturation plateau, the excess sites compete with one another for factor molecules. As the site concentration increases in the nuclei, these factors progressively repartition from the non-specific sites with which they are reversibly associated to the added specific sites, according to the particular values of the partition function Kr that obtain for each factor in vivo (Kr is the ratio of equilibrium constants for reaction of a given factor with its specific sites to that for reaction with nonspecific DNA sites). Cooperative binding amongst the many different factors required for productive expression of Cyllla · CAT, or additional stabilization of active transcription complexes, is also implied. Otherwise, as excess copies of the regulatory domain are introduced, the factors would distribute independently amongst them. The number of completely loaded regulatory domains would thus decrease, and a decline in activity rather than the saturation plateau observed convincingly by Livant et al. (1988) would result.
The experiments shown in Fig. 2 confirm the evidence of Livant et al. (1988) that when the whole regulatory domain is introduced in molar excess with respect to saturating levels of Cyllla · CAT, competition for necessary factors occurs in vivo. This is reflected by an almost stoichiometric decrease in CAT expression, as a function of competitor:CyIIIa · CAT molar ratio. When similar competition is observed following introduction of molar excesses of a given site, as in the experiments of Fig. 3, these competitor sites therefore must effectively titrate away the factor species that recognizes them specifically. However, we have not shown explicitly that the competitive effects are direct. For example the striking P5 competition shown in Fig. 3C could be an indirect result, were the sequestration of the P5 factor actually to affect some regulatory gene, that in turn produces some other factor that affects Cyllla · CAT expression. The improbability of such an interpretation depends on the frequency with which factors required for regulation of the Cyllla gene are also required for regulation of genes whose products are necessary for Cyllla expression, e.g. these same factors, or polymerase components, aspects of processing mechanism, etc. Needless to say, that frequency cannot yet be assessed. But, it is worth noting that in two cases where there is relevant deletion evidence, the same result is obtained by both deletion and competition methods. Flytzanis et al. (1987) showed that severe depression of Cyllla · CAT expression results from excision of the P8 sites (their ΔSP, ΔSPP deletions), or of the P5 site (their ΔH deletion, in which exactly the same fragment was removed from Cyllla · CAT as was used for the experiments of Fig. 3C). It follows that the strong competitions shown for the P5 and P8 sites in Fig. 3B and 3C are indeed direct rather than indirect effects. More generally, however, it is useful to recognize that excess site competition, and site deletion followed by gene transfer, are complementary rather than redundant ways of probing the significance of regulatory interactions in vivo. Deletion does indeed reveal site function directly, but in the context of more or less disturbed cis- regulatory DNA structure. While this may usually be of secondary or no importance, interaction between bound factors could depend in given cases on regulatory domain structure. In one example, a cardiac myosin fight chain gene in which the relevant sites are closely apposed, intranuclear competitive titration yielded consistent results where deletion did not (Braun et al. 1989). Recently competitive titration has been used in studies of the regulatory molecular biology of sea urchin embryo histone genes (Lai et al. 1989), human actin genes (Muscat et al. 1988) and a gut-specific esterase gene of C. elegans (McGhee, personal communication).
Identification of positively acting Cyllla regulatory sites
The surprising result from the experiments shown in Fig. 3 and Table 1 is that out of ten subfragments tested five include sites that compete effectively at low molar excess with respect to Cyllla · CAT expression. These are, in order of their competitive potency, the subfragments containing the P4, P8(I+II+III), P5, P7I and P2(I+II) sites. In addition, the Pl factor may compete weakly and P3B octamer factor is clearly necessary, though its role in Cyllla gene regulation remains unspecified, since its sequestration in a minor fraction of cells is lethal to the embryos. The positively acting factors listed here are probably entirely different and independent of one another, since the sequences to which they bind are wholly distinct (Thézé et al. 1990), and in vitro measurements indicate that these factors are of different prevalences (Calzone et al. 1988). Furthermore, these particular factors accumulate in the embryo with different kinetics (Calzone et al. 1988), and interact with different unique sets of genes (Thie-baud et al. 1990). The three most potent competitors are those subfragments bearing sites for factors P4, P8(I+II+III), and P5. In all of these cases, the individual subfragment competes within a factor of two as efficiently as does the whole regulatory domain. Considering the statistical nature of these observations, and the failure of the whole domain to compete ideally (α=0.76; Fig. 2), it seems reasonable to conclude that the competitive efficiencies of these subfragments cannot be distinguished from that of the whole regulatory domain. Thus, each of these subfragments must be the site of absolutely required, positive interactions, without which Cyllla · CAT activity is minimal. The prevalence of the P4, P8 and P5 factors in the embryo nuclei is close to that of the Cyllla · CAT gene introduced in our competition experiments (i.e. ⩾500 molecules/ nucleus of P4 and P5, and about 1500 of the P8 factors). This prevalence will be rapidly exceeded as molar excesses of competitor sites are introduced. P4 is a CCAAT factor and the essentiality of this interaction is thus not surprising, though it may be noted that there are probably several different CCAAT binding factors in sea urchin embryos (Barberis et al. 1987). Although molar excesses of sites should also have been attained in vivo for the relatively rare P7I factor (⩾200 copies/ nucleus) and for Pl (⩾1100 copies/nucleus), competition in these cases is-only around one-fourth to one-sixth as efficient as with the whole domain. Unless the Pl and P7I factors are distributed to only a small subset of cells, in which their concentrations per nucleus would be 5–10× higher, it can be concluded that these interactions are not individually absolutely required for Cyllla’CAT function, as are those mediated by the sites on the P4, P8 and P5 subfragments. Yet they are necessary for maximum activity.
There can be two reasons for failure of competition when competition sites are introduced: factors with relatively low Kr will not partition to the added sites efficiently; and for factors that are very prevalent it is more difficult to introduce a true excess of specific sites. High endogenous factor prevalence (⩾6000 copies/ nucleus) could easily explain the relatively weak competition-displayed by the P2 subfragment. The number of Cyllla · CAT genes plus competitor would only begin to exceed the number of P2I and P2II factor molecules at the highest molar excesses utilized (cf. Fig. 3E). Thus it cannot be excluded that one or both of the P2 factors is not as essential as at least one of the P8 and the P4 and P5 factors; that is, that the efficiency of P2 subfragment competition is depressed to about one-fifth of the competition efficiency of the whole domain only because of failure to achieve site/factor ratios equivalent to those in the other experiments shown. It is not feasible to introduce greater masses of DNA without incurring risk of embryo lethality irrespective of the sequence injected (McMahon et al. 1985). On the other hand, the almost complete failure of the P3A, P6 and P7II subfragments to compete effectively cannot be explained simply in the same way. The average numbers of molecules of these factors per nucleus are not very different than those for which clear competition was obtained (P3A, ⩾720 molecules/nucleus; P6, ⩾2800 molecules/nucleus; P7II, ⩾1480 molecules/ nucleus; Calzone et al. 1988). Though the Kr values of these factors differ (Calzone et al. 1988), they are all sufficiently high so that competition should have been observed at the molar ratios actually achieved, were these positively acting factors required for embryonic Cyllla · CAT expression.
The Cyllla regulatory domain is a logic switch
Interference with the interactions so far discussed depresses Cyllla · CAT function in vivo, but in all cases the residual expression is spatially correct. As shown in the accompanying paper (Hough-Evans et al. 1990), in embryos with reduced Cyllla · CAT function due to in vivo titration, CAT mRNA is still confined to aboral ectoderm cells. While these interactions are all positive in sense, the P3A and P7II subfragments, which do not competitively titrate down Cyllla · CAT activity (Fig. 3 and Table 1), function in vivo as negative regulators of spatial expression. Viewed as a functional unit, the Cyllla regulatory domain can be seen to require that a number of distinct factor-site interactions take place in order for the gene to function properly in time, in space (i.e. cell lineage) and in amplitude. The inference from what we are learning about this developmental gene regulatory system is that the multiple interactions it requires are not, as may appear superficially, redundant in a functional sense. They exist to carry out diverse biological functions. Perhaps each bears a specific functional relation to the biological status of the cell, that is, to developmental stage, state of specification or lineage, etc. In the case of the P8 and P5 factors, for example, which seem at first sight to function redundantly, i.e. as essentially positive regulators (this paper), we have good reason to suspect quite different biological roles. The P8 factors are present in the embryo before the Cyllla or Cyllla · CAT genes are activated, and their concentration remains unchanged as development proceeds (Calzone et al. 1988). Thus they are unlikely to be responsible for temporal activation, but to be needed as amplitude controllers. On the other hand, the same study shows that the P5 factor rises dramatically over the same period in which this and other aboral ectoderm genes are turned on (Davidson, 1989). P5 could control this important temporal developmental function, since it too is essential for expression, and this factor is probably utilized in the regulatory domains of several other genes that are activated at the same stage of development (Thiebaud et al. 1990). Thus we think of the Cyllla regulatory system as a developmental logic switch, in the sense that it converts the multielement pattern of individual, functionally diverse interactions within it, into a single scalar function, viz. increase (or decrease) in the rate of transcriptional expression.
ACKNOWLEDGEMENTS
This research was supported by NIH Grant HD-05753. R.A. is supported by an NIH Training Grant (GM-07616).