Gene regulatory factors of the sea urchin embryo: II. Two dissimilar proteins, P3A1 and P3A2, bind to the same target sites that are required for early territorial gene expression

As described in the preceding paper (Calzone et al.1991), DNA-protein interactions at cis-regulatory target sites called the P3A sites are required for spatial regulation of the Cyllla actin gene. These interactions cause repression of the CyIIIa gene in territories other than the aboral ectoderm, to which CyIIIa expression is normally confined (Hough-Evans et al. 1990). Calzone et al. (1991) purified by affinity chromatography a regulatory factor, P3A2, which binds tightly and specifically to the P3A target sites. In this paper, we describe a second DNA-binding factor, P3A1, which recognizes the same target sites. P3A1 was isolated by direct ligand screening of a cDNA library made from cleavage stage embryo mRNA. On cloning and sequencing P3A1, we discovered, unexpectedly, that the early embryo in fact contains two very different DNA-binding proteins with high target specificity for the same P3A target site sequences. P3A1 is of the finger protein family, most closely related to the Drosophila Zn finger factor that is the product of the hunchback gene. P3A2 bears almost no resemblance to P3A1 in sequence, except for two short, adjacent sequence motifs that lie within the DNA-binding domain of P3A2 and also appear within the finger regions of P3A1. Though the P3A2 sequence does not possess the canonical features of known classes of finger proteins, these similarities in sequence may indicate the locus of its target site specificity.

There are several curious and interesting features of the P3A regulatory system. First, it appears that wholly dissimilar proteins have evolved to recognize the same regulatory target sequences. This in turn implies a functional significance, e.g. competitive interaction of P3A1 and P3A2 at these sites. In addition, the P3A factors must interact with other regulatory elements to account for the fact that they are required for specification of gene expression in nonoverlapping and functionally distinct lineage elements of the early embryo. That is, in the aboral ectoderm the CyIIIa gene is on and the SM50 gene is off, while the converse is true in skeletogenic mesenchyme, and yet both genes may require P3A interactions to prevent ectopic expression.

Total RNA was isolated from unfertilized eggs or embryos, as described by Posakony et al. (1983) and Lee et al. (1986). Poly(A)⁺ RNA was partially purified from the total RNA, as described by Posakony et al. (1983). 5μg of poly(A)⁺ RNA prepared from embryos of different stages were used to synthesize cDNA. The BRL cDNA synthesis kit was used with oligo(dT) primers. The cDNA was prepared for cloning as described by Watson and Jackson (1985) and ligated into λgtll (Promega Biotec) which had been digested with EcoRl and dephosphorylated, or into λZAP DNA (Stratagene) Ligated DNA was packaged in vitro using a commercial extract (‘Gigapack Gold’, Stratagene). The cDNA library made in λgtll from poly(A)⁺ RNA prepared from 4H (3rd-4th cleavage) embryos contained 5.2×10⁶ primary plaques, and was not amplified before use (average insert size was 2 3kb). The second cDNA library was constructed in λZAP from poly(A)⁺ RNA prepared from 14 h (9th cleavage) embryos. This library contained 8 1×10⁵ primary plaques and was amplified to a titer of l×10¹¹ pfu ml⁻¹.

An aliquot of the 4H embryo λgtll cDNA library was probed by the direct ligand filter detection method of Vinson et al (1987) 1 3×10⁶ plaques were screened at a density of 2×10⁴ plaques per filter, after lytic infection of E. coli strain Y1090r^-(Young and Davis, 1983) The probes used in screening the filters were prepared by nick translation (Rigby et al 1987) of size-selected, double-stranded concatenated oligonucleotides 25/26 (see Calzone et al. 1991 for sequence) using α-³²P-dATP and α-³²P-dCTP (800 Ci mmol⁻¹), according to Vinson et al. (1988) Typical labeling reactions yielded probe with a specific activity of about 5×10 ⁸ ctsmm⁻¹ μg⁻¹ In the initial screen, the filters were exposed to 2×10⁶ctsmin⁻¹ ml⁻¹ of radio-labeled P3A probe DNA. 45 positive clones were identified and one λ38 was found to bind the P3A DNA probe in the following rescreens. Lambda 38 phage DNA was purified in large scale by standard methods A 1.5 kb cDNA insert was subcloned into the EcoRl site of pBluescript SK⁻ (Stratagene) creating clone p38. A fragment of p38 was nick translated and used to screen 6×10⁵ ZAP plaques from the 14 h library. 112 positives were found, 50 of which were plaque purified and converted into plasmids according to the manufacturer’s protocol (Stratagene). The longest clone, p43, was isolated for further analysis. An oligonucleotide was synthesized complementary to the 5’ region of p43 (5’-AATGCAGAAACATCTCCAGGGTCATATAGATCGAG, position 127 – 161 in Fig. 4) and used to rescreen the same ZAP library. The longest clone isolated in this screen, p26, was selected. Another clone, pl3, extended furthest in the 5’ direction, and was used to obtain the 5’ terminal sequence as indicated in Fig 4.

The mRNA coding for the P3A1 factor was later shown by single strand probe excess titration to belong to the rare class of maternal and early embryo mRNAs (Cutting et al. 1990). At 4H postfertilization (the stage from which λgt11 cDNA library was made), there are only about 600 copies of this mRNA per embryo Since there is about 30 pg of total mRNA in the egg, of average length ∼2.5kb (Davidson, 1986, p.71), P3A1 message constitutes only 3×10⁻⁵ of the mRNA, and we should have expected on the average 37 P3A1 clones to have been present in the 1.3 × 10⁶ clones initially screened. Of these about six should have been in the correct orientation and reading frame. That we found only one such clone could be due either to failure of many clones to include the 5’ half of the mRNA, which codes for the necesssary DNA-binding regions of the protein (see below); to insufficiently vigorous protein expression, or to a bias against cloning of this particular sequence. However, screening by DNA hybridization of other various embryo λ libraries with P3A1 DNA probes demonstrate that if anything this sequence is slightly over-represented. A subsequent direct comparison with a clone encoding a different DNA-binding protein isolated from the same library, of the number of clones that tested positive by direct DNA ligand screening with those positive by nucleic acid homology, yielded a ratio of 1:46 Taking into account the requirement for correct orientation and reading frame, the ratio of clones expected to be recovered by direct DNA ligand screening to clones actually detected was 7.7. We were thus fortunate to have recovered the single initial P3A1 isolate. The moral of this excursion is that for this method to be routinely useful in cases where the factor is encoded by an mRNA of such low prevalence, very large numbers of cDNA clones must be screened.

The P3A1 factor is present in the nuclear extracts from which P3A2 was isolated by affinity chromatography, according to reaction of the anti-P3Al antibody. However, P3A1 did not bind to this column, perhaps because in the nuclear extracts the P3A1 binding site is occluded by another protein bound to it. Thus it was fortuitous that we applied both direct ligand screening and affinity chromatography to the problem of isolating the P3A factor(s)

Clones p38, p43, p26, p13 and some additional overlapping cDNA clones were sequenced on both strands using subclones generated by an exonucleaseIII/Sl kit (Promega Biotec), using XL1 blue cells as host (Henikoff, 1984). Clones with deletions of the appropriate size were identified by agarose gel electrophoresis and used for dideoxy sequencing (Sanger et al. 1977) using Sequenase (USB) after alkali denaturation of the DNA (Mierendorf and Pfeffer, 1987). Gaps or ambiguities in the DNA sequence were resolved by using internal primers. The sequence data were compiled using the IBI program. The 5’ end of the insert in the original isolate, λ38, was sequenced using a λgtll primer (New England Biolabs).

An XhoI-BamHl fragment was released from p43 (position 52 – 2473) and end filled. A BamHI linker was used to create the correct frame for ligation into the expression vector pET3c (Studier and Moffat, 1986) generating clone pET3c-P3A1. A second fragment released by digestion with Hindi (position 332-846) was ligated in a similar way into pET3a generating clone pET3a-P3Al. BL21(DE3)pLysS cells were infected with these constructs or, with pET3c vector alone, grown to mid-log phase, induced with IPTG for 2H and harvested. For electrophoretic sizing and subsequent ligand gel blots, cells were lysed in SDS sample buffer and urea was added to a final concentration of 4M. The samples were then heated for 5 nun at 90°C and the solubilized protein was loaded onto SDS-polyacrylamide gels (Laemmh, 1970). For DNAase I footprint and gel shift analyses, the cells were harvested and resuspended in buffer C (20 mM Hepes [pH.7.4], 40mM KC1, 0.1 HIM EDTA, ImM DTT 20% glycerol) with the addition of lysozyme [lmgmF¹]). Cells were incubated on ice for 1 h and then briefly sonicated. The insoluble material was spun out and the supernatant was stored at -70°C For DNAase I footprinting and gel retardation experiments bacterial extracts of P3A1 expression clones were prepared as described by Calzone et al. (1991). The recombinant protein produced by pET3c-P3Al was purified by affinity chromatogTaphy using a concatenated DNA-binding site derived from the Cyllla gene (ohgonucleotides 11/12 of Th6z6 et al 1990).

Bacterial proteins produced by the pET expression construct were resolved on SDS-polyacrylamide gels and electrophoretically transferred to nitrocellulose paper, using standard blotting conditions. The filters were then probed according to the protocol of Miskimins et al (1985) using the same buffer system as in the ligand clone screening method of Vinson et al (1988). The four double-stranded DNA probes used represented binding sites found in the upstream region of the Cyllla gene (Th6z6 et al 1990) and were constructed from complementary ohgonucleotides with offset termini to generate ‘sticky ends’ that would allow concatenation. The sequences of some of these probes are given in Table 1 of Th6z6 et al. (1990). Those utilized in this study were ohgonucleotides 23/24 (the PI site); 5/6 (the P6 site); 19/20 (the P7I site), all given by Th6z6 et al. (1990) Ohgonucleotides 25/26 are given in the accompanying paper (Calzone et al 1991), and the relevant native P3A site of the SM50 gene is shown in Fig. 3C of this paper (see Table 2 of accompanying paper for other P3A target sites in these genes)

The quantitative gel shift reactions used for estimation of K^, were carried out as described by Calzone et al. (1988) and in legend to Fig. 2. DNAase I footprints were carried out by standard methods as described in detail by Calzone et al. (1991).

A Bluescnpt cDNA clone containing the complete coding sequence of the P3A2 protein (Calzone et al. 1991) was used as the starting material. The 3’ series of deletions was generated from a slightly shortened version of this clone, in which 30 nt were removed from the 5’ end by exonuclease digestion (to eliminate a foldback that caused premature stops in transcription). The shortened clone, pP3A2A, was linearized by digestion with different restriction enzymes, at the following positions in the sequence (Calzone et al. 1991):

To construct 5’ deletions PCR was used to generate copies of selected regions of the P3A2 coding sequence (Sarkar and Sommer, 1989). The template was a Hindlll-Htndlll fragment from the P3A2 clone [position 1 – 2883 in the P3A2 DNA sequence (Calzone et al. 1991)]. The ohgonucleotides that were used contained, in addition to a sequence complementary to the desired locations in the P3A2 cDNA, a 23-base T7 promoter sequence followed by a 9-base translation initiation signal that fused a methionine in frame with the P3A2 protein sequence. These oligonucleotide primers were as follows (the digits in the name for each oligonucleotide indicate the start of the region of the cDNA clone to which the oligonucleotide is complementary):

PCR was performed with oligonucleotide 1163 (3’-end), and one of five additional ohgonucleotides (5’-end) using a Perkin-Elmer ‘Gene-Amp’ kit according to the manufacturer’s recommendation in a 100pi\ reaction. After 10 min at 94 °C, 35 cycles of PCR were performed with the Perkin-Elmer Cetus automated thermal cycler. 1 jA (–0.5 fig) of the amplified DNA were added to 43 jA of RNA transcription mixture and was transcribed using T7 RNA polymerase (Stratagene)

The linear DNA obtained from the PCR reactions was used as template for T7 RNA polymerase to transcribe uncapped RNA suitable for in vitro translation (Promega Biotec). The RNA templates (about 2fig) were added to a rabbit reticulocyte lysate (Promega Biotec) containing [³⁵S]methionine according to the manufacturer’s specifications, and the protein products were confirmed by analysis of the reaction products on SDS-polyacrylamide gels.

Target binding sites for the P3A factors in the SM50 and Cyllla genes show a core sequence element, ^C/TX ^C/TGCGC^A/_T (Calzone et al. 1988, and preceding paper; Th6z6 et al. 1990; Thiebaud et al. 1990). The P3A1 factor described herein was isolated by screening an expression cDNA clone library with concatenated oligonucleotides, each monomer of which contains two of these core sequence elements. This probe represents a naturally occurring, high affinity double P3A site located at position -128 to -102 of the SM50 gene. The exact sequence of this site is shown in Table 2 of the accompanying paper of Calzone et al. (1991). Oligonucleotides 25/26, which were used to produce the probe, consist of precisely this sequence with a few additional flanking nucleotides from the SM50 gene, as also given in the accompanying paper. The concatenated probe was labeled and used to screen a set of early embryo cDNA libraries by the method of Vinson et al. (1988). The initial P3A1 isolate was obtained from a Agtll library containing inserts of 4h embryo cDNA. One reproducibly positive isolate, called A38, was obtained from 1.3x10 plaques produced by a portion of the unamplified packaging mixture. After purification the insert of this recombinant, which was about 1.5 kb in length, was converted into a subclone (p38) and a fragment of p38 was used to screen a second cDNA library, produced from 14 h embryo cDNA, and constructed in AZAP. The longest of the positive clones contained a 2.6 kb insert (p43), and this clone was used for much of the following study.

An insert from clone p43 containing sequence coding for amino acids 9– 387 (the -COOH terminus) was transferred to the pET3c expression vector (Studier and Moffat, 1986). Tliis clone is denoted pET3c-P3Al. After IPTG induction the proteins were extracted (see Materials and methods, and legend to Fig. 1), and loaded onto an SDS-polyacrylamide gel. Following electrophoresis they were transferred to nitrocellulose and reacted with the probe containing the P3A site, and with three other probes that bind tightly to other sea urchin embryo nuclear factors involved in the regulation of the CyIIIa gene, viz the Pl, P6 and P71 sites (Calzone et al. 1988; Thézé et al. 1990). Fig. 1 shows that only the P3A probe is bound by the bacterial products of the expression clone pET3c-P3Al. This result is dependent on IPTG induction, and is not obtained with extracts from bacteria containing the vector alone (data not shown). The apparent relative molecular mass of the active fusion protein is about 60×10³ (in this vector the fusion product includes only 12 amino acids encoded by the host sequence).

The fusion protein produced by the pET3c-P3Al expression construct was recovered and used for quantitative gel shift reactions in order to measure its relative affinity for the SM50 and CyIIIa target sites. The protein product of pET3c-P3Al was purified using a site-specific DNA affinity column (Calzone et al. 1991). We utilized quantitative gel shift competition series (cf. Calzone et al. 1988) to determine values for the equilibrium dissociation constant (K_D) by least squares analysis. Representative gel shift competitions are shown in Fig. 2A, for reaction of the recombinant P3A1 protein with the CyIIIa target sequence; and in Fig. 2B for its reaction with the SM50 target sequence. Beneath each panel the data are plotted in a form which allows K_D to be calculated directly from the slope (which, as given in the legend, is –K_D⁻¹). Fig. 2 shows that the bacterial protein produced by pET3c-P3Al indeed reacts specifically with both the SM50 P3A and the CyIIIa P3A sites. The K_D values obtained in these particular reactions are about the same, i.e. 6.0× 10⁻⁸ M for the CyIIIa target fragment and 3.5×10^{− 8}M for the SM50 target fragment. These values are not significantly different, and experiments in which the SM50 oligonucleotide competitor was used with the Cyllla probe, show that the recombinant P3A1 protein partitions equally between these target sites. It is useful to compare these values with the K_D measurements obtained in similar experiments using the P3A2 factor, and presented in the accompanying paper of Calzone et al. In contrast to the equivalent binding affinities for the SM50 and CyIIIa sites displayed by P3A1, P3A2 binds to the SM50 probe about 50-fold better than it does to a probe containing a single CyIIIa site, though it binds to the Cyllla site with about the same affinity as does P3A1. Thus the apparent K_D values extracted from the experiments of Calzone et al. were K_D=7.4×10⁻⁸M for the reaction of P3A2 with the CyIIIa site, but K_D=1.7×10⁻⁹M for its reaction with the SM50 probe. However, there are two reasons why interpretation of this comparison in itself is not straightforward, (i) The Cyllla fragment used for the gel shifts as probe contains a single target site while the SM50 probe fragment contains two nearby but nonidentical sites (see below), and (ii) the P3A1 protein was a recombinant product of a bacterial expression system that was affinity purified, while the P3A2 protein utilized for these studies was purified directly by affinity chromatography from sea urchin embryo nuclear extract, and this factor is subject to modifications that may affect its DNA-binding activity. Nevertheless, the low molar values for K_Dmeasured in Fig. 2 of this paper, which fall well within the range reported for known DNA-binding regulatory proteins, demonstrate that P3A1 is indeed a specific DNA-binding protein that recognizes the P3A target sites of both SM50 and CyIIIa genes. Furthermore, although proof of the point must await direct purification of the native P3A1 protein, the comparison with P3A2 raises the possibility that P3A1 does not discriminate between the CyIIIa and SM50 sites, while P3A2 does. The same distinction obtains in the case of closely spaced P3A sites in a third gene, Sped. Here again P3A2 binds these sites relatively tightly, but P3A1 responds to them the same as it does to the single CyIIIa site utilized for the experiment of Fig. 2A (data not shown). Were this difference to obtain in vivo. differential functional significance of the two factors would be implied, with respect to interactions with the CyIIIa, SM50 and other genes that bear regulatory P3A target sites.

The footprints and hypersensitive sites shown in Fig. 3A,B demonstrate explicitly the sequences engaged by the DNA-binding domains of the P3A1 factor in the SM50 and Cyllla target sites. For these experiments, a second P3A1 expression clone was used (pET3a-P3AlD), which contains a shorter insert coding for amino acids 104 – 272. The protein encoded by this construct includes the complete DNA-binding domain of the P3A1 protein. The truncated pET3a-P3A1Δ protein was employed for the experiments of Fig. 3 because its accumulation in the bacterial expression system and stability in extracts greatly exceeds that of the nearly complete protein encoded by pET3c-P3Al.

In Fig. 3C the target regulatory sequences of the CyIIIa and SM50 genes that are protected by P3A1 are compared with those protected by P3A2. P3A2 footprints are shown in Calzone et al. (1991) and in B of Fig. 3. Both proteins clearly protect the core ^T/_CX ^T/_CGCGC^A/_T elements of these sites, which are marked by boxes in Fig. 3C (their relative orientations are indicated by arrows). Methylation interference experiments (not shown) also indicate that both proteins make contacts with G residues on both strands of the central GCGC element. These data, the footprint studies summarized in Fig. 3C and the oligonucleotide competition results shown in Fig. 2 confirm that P3A1 and P3A2 react with the same DNA target sites. On the CyIIIa site the DNAase I footprint patterns overlap exactly on the 5’ → 3’ (top) strand, and within one or two nucleotides at each end on the 3’ ← 5’ (bottom) strand. However, the P3A1 protein binds very differently from the P3A2 protein on the double SM50 site. P3A1 reacts strongly with only the proximal site, where it covers 13 nucleotides just as on the lower strand of the Cyllla site, while P3A2 equally protects both sites (this almost certainly requires two P3A2 molecules). At higher concentrations P3A1 weakly interacts with the upstream SM50 site as well. This impressive difference between the P3A1 and P3A2 footprints obviously could account for the striking difference in the measured K_D values for the two proteIIIs. Thus the footprints provide a reasonable qualitative interpretation for the observation that P3A2 prefers the SM50 probe about 50 times more than it does the CyIIIa probe, while P3A1 regards them more or less indiscriminately, and binds to both with about the same affinity as P3A2 does to the CyIIIa site. However, the P3A1 protein may respond specifically to a different feature of the proximal SM50 site, viz that if considered by itself, this site is symmetrical (vertical arrow). Thus the P3A1 factor positions itself directly over its axis of symmetry, so that both termini of the protected region lie in equivalent positions, within the sequence elements C^A/_GCCCC. This feature of the SM50 site could be functionally important, since this symmetrical sequence motif, to which P3A1 is sensitive, is not found in the Cyllla site.

Fig. 4 displays the sequence of the P3A1 cDNA and the protein for which it codes. The actual mRNA probably begins a short distance upstream of the 5’ terminus shown. However, it is unlikely that this mRNA has a very large 5’ untranslated leader because, though the clone includes only 25 nucleotides prior to the translation initiation site, the size of the sequenced transcript, 2473 nt, is indistinguishable from the size of the transcript revealed by RNA gel blots (Cutting et al. 1990) measured by reference to standards as about 2500±50nt. Use of the first ATG is supported by the adjacent sequence CÂAÂTCATGG, which at all except the starred positions is identical with the 10 nt consensus translation start site of Kozak (1987). There follows an 1161 nt open reading frame, a stop codon at position 1186 in the sequence, and a 1287 nt long 3’ untranslated region, which in the recovered clones lacks a poly (A) tail, though the mRNA is indeed polyadenylated in the embryo (Cutting et al. 1990). The open reading frame of the P3A1 sequence specifies a protein with a relative molecular mass of 42×10³ (387 amino acids), although, as noted above, the recombinant expressed protein migrates on SDS-polyacrylamide gels as if its mass were 6O×lO³ (Fig. 1, lane 1). Similar overestimates of size have been noted for other transcription factors, including GCN4 (Hope and Struhl, 1986), RAP1 (Shore and Nasmyth, 1987) and HSF (Sorger and Pelham, 1988). Since in the present case the recombinant protein was expressed in bacteria, its anomalous migration is unlikely to be due to significant posttranslational modification, and its cause might be sought in its unusual amino acid composition. The P3A1 protein is predicted to have a high content of serine, threonine and proline, together ∽29%, as also noted for some other DNA-binding proteins, e.g. GF1 (Tsai et al. 1989), Spl (Kadonaga et al. 1987) and SRF (Norman et al. 1988). The amino acid terminal region of the P3A1 protein (positions 1–53) is strongly acidic, with a net charge of –11 and perhaps could function as a transcriptional activator region (reviewed by Mitchell and Tjian, 1989). A short strongly basic region (7/11 amino acids) appears after position 237, similar to the nuclear localization signals described in other proteins (Dingwall and Laskey, 1986). No significant similarities are detected, comparing the derived P3A1 protein sequence with those in the available sequence data banks, except for two regions that are obviously similar to the Zn finger sequences of many known DNA-binding proteins. These regions are indicated by shading in Fig. 4. They conform to the general form of Zn finger motifs, Phe-X-Cys-X_(2–4)-Cys-X₃-Phe-X₅-Leu-X₂-His-X_(3–5)-)-His [as described by Miller et al. (1985) and Berg (1988)]. We showed in the experiment of Fig. 3 that a truncated version of the P3A1 protein including only amino acids 104-274 and thus these two finger sequence elements, is capable of binding the P3A target sites specifically.

In Fig. 5 the P3A1 finger region is shown aligned with a number of other finger proteins, all of which are known directly to serve as DNA-binding regulatory factors, so as to display the subfamilial relationships among the examples included (see legend). The invariant C-C and H-H residues (bold face in Fig. 5), as well as several other residues that frequently (i.e. in over half the examples) also occur in specific locations in Zn finger domains (boxed residues) are also present in the P3A1 finger sequences. There is no evidence for P3A1 that Zn²⁺ or other metals are required for DNA binding, however. Fig. 5 displays a specially close relationship between the sequence of the P3A1 fingers and those of the Drosophila hunchback factor. Amino acids shared (in register) between either of the P3A1 finger sequences and any of the other finger sequences in Fig. 5 are shaded (excluding those that are common to more than half of all the Zn fingers shown, i.e. the invariant C and H residues and the boxed residues). It can be seen that of the six hunchback fingers two share six such residues and two five with the P3A1 finger sequences. One finger sequence of the four snail fingers also shares five such residues, while all other fingers in Fig. 5 are less similar. Furthermore, other amino acids that are common (circled) in the 18 finger regions excluding the hunchback and P3A1 fingers are rarely seen in the latter two sets of fingers. For example, five of the six hunchback fingers lack the aromatic tryptophan or phenylalanine residue that occurs four amino acids beyond the second conserved cysteine and that is present in all 18 of the other fingers shown in Fig. 5. Berg (1988) suggested that this aromatic amino acid is part of a hydrophobic core important for the folding of zinc fingers and that this amino acid can be replaced by another aromatic residue positioned two amino acids following the second cysteine; indeed 50% of both the hunchback fingers and the P3A1 fingers display tryptophan or phenylalanine at this position, while these amino acids are found in none of the other 18 finger structures included in Fig. 5.

Standard comparison of the P3A1 and P3A2 sequences indicated that the proteins are almost entirely dissimilar (only about 11% identical amino acids). This result poses something of a paradox, since as shown in Fig. 3 the two factors bind to essentially the same target DNA sequences. To obtain a more focussed insight, we produced N- and C-terminal deletions of the P3A2 protein, and determined their effect on its DNA-binding activity, since the nature of the P3A2 DNA-binding domain was not obvious a priori from any features of its primary sequence. A cDNA clone containing a complete coding sequence, for P3A2 (Calzone et al. 1991) was truncated at its 3’ end at conveniently located restriction sites, or at its 5’ end by oligonucleotide primed PCR mutagenesis, as described in Materials and methods. The truncated clones were then transcribed using the vector T7 polymerase start site, and the RNAs were translated in vitro. The series of truncated products resulting from the N-terminal and C-terminal deletions are shown respectively in Fig. 6a1 and b1. Gel shift reactions with the SM50 target site were then carried out with these polypeptides as shown in parts a2 and B2 of Fig. 6. At the N terminus, the P3A2 region required for DNA binding evidently includes a sequence that lies between amino acids 25 and 90, since of the N-terminal deletions, only that terminating at position 25 still recognizes the P3A probe [Fig. 6a2], Deletions to amino acid 90 and all further deletions completely abolish binding. Truncation from the C-terminal end showed that a P3A2 polypeptide including residues 1-258 exhibits normal binding [Fig. 6b2], while further deletion to position 222 abolishes specific DNA binding. The DNA-binding domain of P3A2 is thus <233 and >130 amino acids in length and lies between positions 25-90 and 222 – 258. Though of low resolution, this delimitation of the P3A2 DNA-binding region afforded an interesting observation, shown in Fig. 7A. When carefully examined for sequence similarities with P3A1, five amino acids that lie within or adjacent to the first finger region of P3A1 are also found, in registration, within the DNA-binding region of P3A2. Furthermore, if a gap of four amino acids is allowed between the fingers of P3A1, a similar feature is revealed in the second finger of P3A1, where there are four amino acids in the DNA-binding region of P3A2 that are identical, in register, with respect to this P3A1 finger. In both fingers, the identical amino acids begin with the aromatic ammo acid following the second cysteine. However, as Fig. 7B shows, the overall sequences of P3A1 and P3A2 show no other significant similarities whatsoever, even when the alignment is chosen on the basis of these nine identical finger region residues. Since the two factors recognize almost the same target sequences, these observations invite a direct test by mutagenesis of the significance of these nine amino acids for the target specificity of the P3A2 protein.

The P3A1 target site appears in the regulatory domains of at least three territory-specific genes that are activated regionally in the early sea urchin embryo and, in one of these genes, interactions at this site are known to be required for the correct spatial pattern of expression. In this paper, we describe the P3A1 DNA-binding factor, which specifically interacts with these target sites. Unexpectedly, a second factor, P3A2, that recognizes the same target sites, has also been discovered, as described in the first paper of this series. As we demonstrate in Fig. 7, except for small sequence elements within the DNA-binding regions of these two proteins, the P3A1 and P3A2 mRNAs code for wholly dissimilar polypeptides. P3A1 displays canonical features of the Zn-finger class of DNA-binding protein, while P3A2 resembles no known DNA-binding proteins, and thus appears to represent a novel class of regulatory factors (Calzone et al. 1991).

P3A1 evidently contains two adjacent finger sequence elements of the C-C/H-H class, as shown explicitly in Fig. 5. This structure was initially recognized in the repetitive DNA-binding domain of the TFIIIA transcription factor (Miller et al 1985), in which each of the repeat elements is organized into a finger-like structure by a tetrahedrally coordinated Zn ion at its base. Many DNA-binding proteins utilizing this canonical structure are now known, and a variety of structural and mutational evidence demonstrates that the finger structure is indeed that which mediates the sequencespecific interactions with DNA target sequences in which these proteins engage functionally (e.g. Diakun et al. 1986; Parraga et al. 1988; Nagai et al. 1988; Kadonaga et al. 1987; Berg, 1988; Redemann et al. 1988) . Thus it is not surprising that a truncated 24 × 10³ version of the P3A1 protein that includes its two finger sequence elements displays the same site specificity as does the complete 43×10³ protein.

An interesting subdivision of the class of finger proteins is shown in Fig. 5, where it can be seen that the P3A1 fingers share a number of sequence features with the fingers of the hunchback factor and, like the hunchback factor, lack certain residues that are otherwise common amongst DNA-binding Zn finger protein sequences. However, the target site for the hunchback factor, ^G/_C^A/_CATAAAAAA (StanojeviČ et al. 1989) displays no similarity whatsoever with the target site for the P3A1 factor, ^c/_TX^c/_TGCGC^A/_T. Thus the specific sequences that are responsible for target site recognition within the finger region must include amino acids besides those that are shared specifically between the hunchback and P3A1 fingers. This result is consistent with the observation that none of the hunchback fingers displays the sequence of either of the P3A1 fingers, including the five and four amino acids respectively present in both the P3A1 fingers and the P3A2 DNA-binding domains. Many though not all of these amino acids are located near the tip of the P3A1 fingers, and between the tip and the histidine residues, i.e. in the region believed to be configured into an rr-helix that lies in the major groove of the target site DNA, where sequence-specific contacts probably occur (Lee et al. 1989; Berg, 1988; Gibson et al. 1988; Brown and Argos, 1986). These local sequences might indeed contribute to the target site recognition elements of P3A2 as well as of P3A1, accounting for their almost identical site specificity. Since P3A2 lacks all the standard features of Zn fingers, it would then follow that there must be alternative ways to organize a three-dimensional structure in which almost the same DNA-protein contacts are formed, that do not require the recognition elements to be mounted on a canonical C-C/H-H Zn finger. However, it is important to keep in mind that a much larger domain of the P3A2 factor is apparently required for DNA binding than is included in the two finger structures of P3A1. The deletion experiments shown in Fig. 6 indicate the minimum necessary DNA-binding domain of P3A2 to encompass >140 amino acids. Of course this could include sequences required for other functions interspersed amongst the elements that immediately contact the DNA target site. Since no DNA-binding proteins displaying significant sequence similarities to P3A2 have been reported, much more information will be required to develop an image of structure-function relationships within this large region of the P3A2 factor. The interesting fact remains that a finger protein and a non-finger protein recognize the same regulatory sites with nearly the same affinity.

Specification of both the aboral ectoderm and the skeletogenic mesenchyme, the territories of the embryo in which the CyIIIa and Specl, and the SM50 genes are respectively expressed, occurs during cleavage (Davidson, 1986, 1989). Both P3A1 and P3A2 are encoded by low abundance maternal mRNAs; there are about 300 copies of P3A1 mRNA per egg, rising to about 2000 per embryo at 24 h (at that stage ∽ 4molecules/cell), and about 1100 copies of P3A2 mRNA per egg, a level that remains approximately constant (Cutting et al. 1990). Calculations show that the levels of these mRNAs present in the very early embryo would just suffice to provide sufficient levels of newly synthesized P3A factors in the nuclei of the mid-cleavage stage founder cells, when active P3A factors should first be required (Cutting et al. 1990). Since the amounts of P3A1 and P3A2 mRNA are approximately the same by midcleavage the amounts of the two factors are probably comparable. However, we do not yet know where in the very early embryo the P3A1 and P3A2 factors are located. By late, even mid-embryogenesis, aboral ectoderm is a wholly committed, terminally differentiated cell type, in contrast to oral ectoderm and the vegetal plate derivatives, which continue to differentiate in new directions (Davidson, 1986,1989; Cameron et al. 1987). We suspect that in so far as it pertains to aboral ectoderm gene expression the P3A regulatory system is a specification system that functions very early in development. Thus the factor is present even before the Cyllla gene is activated (Calzone et al. 1988), and competitive interference with P3A2 interactions in vivo results in ectopic specification. These factors could, however, have other functions in the still differentiating embryonic cell types of the oral ectoderm and gutmesenchyme territories. But it seems most unlikely that these maternally encoded factors are physically absent from the aboral ectoderm founder cell lineages in early cleavage. Initially the egg is radially symmetrical in terms of potential to generate oral versus aboral ectoderm. That is, any quadrant of the 4-cell embryo can give rise to a complete larva (Davidson, 1989). An hypothesis (Davidson, 1989) is that the cytological polarization that is the initial indication of this axis causes a modification of radially distributed factors such as P3A1 and P3A2 that would affect their activity along this dimension of the embryonic space.

In considering the P3A regulatory system, we begin with the argument that two different factors that interact with the same or overlapping sites are likely to perform opposite functions, i.e. to act competitively. Since we know that at least one of these factors behaves negatively the other might be required for transcriptional activation. The literature contains several examples of negative factors that probably function by competitively interfering or modifying positive interactions, in which both factors utilize the same or overlapping binding sites. Among such examples are the positive and negative interactions in the interferon gene (Keller and Maniatis, 1988); the CCAAT and CCAAT-displacement factor interactions in a sea urchin histone gene (Barberis et al. 1987); and the overlapping and oppositely acting K_r and bed interactions, and bed and giant interactions identified in the ‘stripe 2’ regulatory element of the Drosophila eve gene (Small et al. 1990). For our case an additional clue is provided by the in vitro interaction studies we present in Figs 2 and 3 of this paper. The most striking difference in DNA-binding activities between P3A1 and P3A2 that we have detected is their response to the target sites in the SM50 regulatory domain. Table 2 of Calzone et al. (1991) shows that P3A2 binds about 50 × more tightly to the double site in this gene than to the single CyIIIa site, and about 10× more tightly to a third target site of the SM50 gene located proximally. In contrast, as shown in this paper, P3A1 recognizes all of the target sites with about the same affinity (Fig. 2), which is also about the same affinity that P3A2 displays for the CyIIIa site. Thus, at relatively low factor concentration, the SM50 target site would be expected to be occupied by P3A2, while the CyIIIa site would not. The two factors could both be involved in regulating both aboral ectoderm and skeletogenic genes if they function competitively. Thus, for example, we may suppose that P3A2 is the factor that functions negatively, as a repressor of transcriptional expression, and that P3A1 competitively antagonizes this repression. Then, were the ratio of P3A1 to P3A2 to increase during development in favor of P3A1 by a modest factor, say over a 10-fold range about equivalence, the transcription of genes such as CyIIIa would be allowed, assuming the remainder of the required transcription factors were available (Thézé et al. 1990; Calzone et al. 1988). However, at such ratios repression of SM50 expression by P3A2 would everywhere be dominant. To account for SM50 activation exclusively in the skeletogenic mesenchyme lineage it is likely, as postulated earlier (Davidson, 1989), that other positive skeletogenic gene activators should be present ab initio in this autonomously specified lineage. A testable prediction is that there should be a much smaller concentration of the P3A2 repressor relative to P3A1 in the confined cellular domain from which the skeletogenic founder cells descend, i.e. the 5th cleavage micromeres, than in any other lineage elements of the cleavage stage embryo.

We thank Mariann Brunkener for construction of the λgtll library. This work was supported by an NIH grant HD-05753. Support for the initial phases were provided by a Developmental Biology grant from the Lucille P. Markey Chantable Trust C.H. was supported by a postdoctoral fellowship from EMBO.