Serendipity (sry) beta (β) and delta (δ) are two finger protein genes resulting from a duplication event. Comparison of their respective protein products shows interspersed blocks of conserved and divergent amino-acid sequences. The most extensively conserved region corresponds to the predicted DNA-binding domain which includes 6 contiguous fingers; no significant sequence conservation is found upstream and downstream of the protein-coding region. We have analysed the evolutionary divergence of the sry β and δ proteins on two separate levels, their embryonic pattern of expression and their DNA-binding properties in vitro and in vivo. By using specific antibodies and transformant lines containing β -galactosidase fusion genes, we show that the sry β and sry δ proteins are maternally inherited and present in embryonic nuclei at the onset of zygotic transcription, suggesting that they are transcription factors involved in this process. Zygotic synthesis of the sry β protein starts during nuclear division cycles 12– 13, prior to cellularisation of the blastoderm, while the zygotic sry δ protein is not detectable before germ band extension (stage 10 embryos). Contrary to sry δ, the zygotic sry β protein constitutes only a minor fraction of the total embryonic protein. The sry δ and δ proteins made in E. coli bind to DNA, with partly overlapping specificities. Their in vivo patterns of binding to DNA, visualised by immunostaining polytene chromosomes, differ both in the number and position of their binding sites. Thus changes in expression pattern and DNA-binding specificity have contributed to the evolution of the sry β and δ genes.

Identification and characterization of an increasing number of proteins regulating gene expression in eucaryotes has led to the discovery that transcription factors make up a remarkably diverse group of proteins constructed from a variety of combinatorially arranged functional domains. Consequently, families of eucaryotic regulators with related DNA-binding specificities and varying levels of functional homology could have been created by gene duplication and subsequent divergence during evolution (reviews by Ptashne, 1988; Mitchell and Tjian, 1989).

The Cys2/His2 zinc finger first identified in the Xenopus transcription factor TFIIIA (Brown et al. 1985 ; Miller et al. 1985) is one well-characterized DNA-binding motif that defines large multigene families, particularly in vertebrates (Bellefroid et al. 1989; Nietfeld et al. 1989). Characteristic of this finger is its tandem repetition occurring a variable number of times per protein (reviews by Evans and Hollenberg, 1988; Payre and Vincent, 1988). The existence of subfamilies of Cys2/His2 finger proteins, characterised by the presence, at their TV-terminal end, of common conserved modules unrelated to the finger, has also been shown in Xenopus (Knöchel et al. 1989). Furthermore, comparison of the gap segmentation gene hunchback between two Drosophila species has revealed that divergence of expression patterns may also contribute to the functional evolution of Cys2/His2 finger protein genes (Treier et al. 1989).

To study structural relative to functional parameters contributing to the evolutionary divergence of Cys2/ His2 finger proteins, we started a few years ago the characterisation of the Drosophila sry β and sry δ genes. These two finger protein genes map at a single locus and probably result from a duplication of an ancestral gene containing at least six fingers (Vincent et al. 1985). Mutants for sry δ, which display a late embryonic lethal phenotype, have been isolated. These mutants can be rescued by transformation with a DNA fragment that contains the sry δ but not the sry β gene (Crozatier et al. manuscript in preparation). Therefore, it is likely that the sry β and sry δ genes are functionally distinct. The physical linkage and evolutionary conservation of the sry β and sry δ coding sequences in distant Drosophila species (Aguadé, 1988; Vincent et al. 1988 and unpublished data) confirms the value of the sry β and δ genes as a potential model for the study of the functional evolution of finger proteins relative to their DNA-binding specificities.

We have recently shown that the sry δ protein is maternally inherited by the embryo and present in nuclei prior to the onset of zygotic transcription. Later in embryogenesis and up to the adult stage, the sry δ protein is present in nuclei of transcriptionally active cells, suggesting that it is a transcription factor with a role in zygotic activation and maintenance of expression of general cellular functions (Payre et al. 1989).

In this paper, we have determined the pattern of embryonic expression of sry β in order to compare it to that of sry δ, by using immunostaining and transgenic approaches that make it possible to distinguish between maternal and zygotic contributions to these patterns. We then show that both proteins made in E. coli display partly overlapping in vitro DNA-binding specificities and in vivo patterns of binding sites on polytene chromosomes.

Plasmid constructions

(a) pSBLl construct

A sry β DNA fragment bordered by a BamHI site 458 bp upstream of the sry β transcription start site, and a SalI site created in the sryβ coding region at codon position 340, after the 6th finger (Vincent et al. 1985), was fused in frame to E. coli lacZ gene using the pSDLl plasmid (Payre et al. 1989) opened with SalI and Xbal. The resulting construct (pSBLl) contained the sry β–lacZ fusion placed upstream of the sry δ polyadenylation site into the plasmid pDm23 (rosy containing P-element transformation vector, a gift of D. Messmer and G. Rubin) derived from the plasmid Carnegie 20 (Rubin and Spradling, 1983). This construct was used to transform flies from the ry506 stock according to standard methods (Rubin and Spradling, 1982). Six independent transformant lines, identified as being different from one another by chromosomal linkage and Southern blot analysis, were made homozygous and are referred to as SBL lines.

(b) pSBAl construct

A sry β DNA fragment, starting at an Xmn1 site at codon position 41 and ending at the SalI site created at codon position 340, was inserted into the plasmid pAR3039 (Rosenberg et al. 1987) to give the pSBA1 plasmid.

The pSDL1 and pSDA1 plasmids are described in Payre et al. (1989).

(c) pRLXSl plasmid

This was generated by inserting a λ fragment from position 31617 to 33499 on the λ map (Daniels et al. 1983) in the plasmid pTZ18R.

Expression in E. coli and purification of the sry β protein

Overexpression of sry β or sry δ in E. coli was achieved by using the pSBAl and pSDAl plasmid as described by Tabor and Richardson (1985). Purification of the inclusion bodies was carried out according to Nagai et al. (1988). After renaturation of the proteins (Hoey and Levine, 1988), extracts were chromatographed on a monoQ HR 5/5 column (Pharmacia) . Fractions corresponding to either sry β or sry δ (purity of at least 90– 95% as estimated by Coomassie staining of PAGE-SDS) were pooled and loaded onto a heparine sepharose column. The fractions eluted at 0.4 M KC1 were dialysed against HK buffer (25mM Hepes, 20% Glycerol, 120 HIM KC1, 1 mM DTT, 50 μm ZnSO4) and used for DNA-binding experiments.

Production of anti-sry β antibodies

20 μ g of pSBAl encoded sry β purified protein mixed together with 4mg of poly(A)-poly(U) in 2ml of NaCl (9%) was injected intravenously into young female rabbits. The injections were repeated five times at ten day intervals. Blood was collected weekly from rabbits boosted every month. Crude sera clarified by a 17000 g centrifugation were passed over a column of affigel 10 sepharose (Biorad) coupled with total bacterial protein from an induced culture of pAR3039. After several passages over the same column the negatively purified serum was loaded onto a similar column containing 2 mg ml-1 coupled purified sry β protein. Following washes in conditions described by Carroll and Scott (1985), the retained fraction containing the affinity purified antibody was eluted by 2 M guanidine HC1 and extensively dialysed against 1× PBS containing 0.1 % sodium azide.

Immunological detections

Western blotting was done according to standard procedures using 1/1000 dilution for both anti-sry β and sry δ and 1β000 for anti-β -gal (Promega). Nuclei were prepared from 2 to 12 h old embryos according to Jowett (1986) and extracted in conditions described by Dignam et al. (1983).

Immunostaining of whole-mount embryos was performed essentially as described by Carroll and Scott (1985). Embryonic stages referred to in the text are those of Campos-Ortega and Hartenstein (1985).

Salivary glands of third instar larvae were dissected in PBS and rapidly transferred to ALF solution (acetic acid 45 %, lactic acid 1:6, formaldehyde 3.7%) for 30min. After fixation, glands were squashed in ALF. Antibody reactions and Giemsa staining were done according to Zink and Paro (1989). Procedures involving the secondary antibody were as indicated by the manufacturer (Vector Lab).

Immunoprecipitation of DNA fragments

In a typical experiment, 20 ng of end labelled DNA and 1.5 μg of yeast tRNA, complemented or not with 200 ng to 1μ g of sonicated calf thymus DNA, were mixed with or without 50– 100 ng of protein in 25μl of HK buffer. Following an incubation of 30 min in wet ice, 10 μ l of 1/10 Staphylococcus aureus cells were first preincubated for 30 min at 4°C with 1 μl of specific antiserum before addition to each sample. After incubation for 30 min at 4 °C, 1 ml HK buffer was added, and samples centrifugated for 5 min at 4°C. After a second wash in 1 ml HK buffer containing 0.5 % NP40, pellets were resuspended in formamide. Immunoprecipitated DNA fragments were then analyzed by electrophoresis on 5 % acrylamide sequencing gels followed by autoradiography.

Photography

Views were taken at varied magnification using bright-field or NOMARSKI optics Nikon (see legends of figures). We used Kodak Ektachrome 50 ASA tungsten or Ektar 25 ASA films for colour photography.

Protein sequence comparison

A preliminary comparison of the sry β and sry δ proteins revealed the presence of a 28– 29 amino-acid motif repeated several times in each protein, the Cys2/His2 finger motif (Miller et al. 1985; Vincent et al. 1985; Vincent, 1986). We extend here the comparison to the entire protein length (Fig. 1), taking into account minor corrections of the published sequence that make the sry β and δ proteins 356 and 434 amino-acids long, respectively (Vincent et al. 1985, see legend of Fig. 1). Along the sequence comparison, the amino-acid numbers refer to the sry β protein sequence. Consequently, the places where the sry δ protein has more amino-acids are considered as insertions, and where it has less as deletions without any implication of directionality in the change seen.

Fig. 1.

Conservation and divergence within the sry β compared to the sry δ protein. The representation is adapted from that chosen in Treier et al. (1989). Conserved residues are indicated by black lines. Deletions/insertions, as defined in the text, are represented by triangles facing downwards and upwards, respectively. The open, and small and large black triangles correspond to 21, 1 and 5 bp insertions, respectively. The following exchanges were considered as conservative ones: L⇌ V⇌ E⇌ D; S⇌ T; Y⇌ F; N⇌ Q. The position of the sry β intron is marked. Open arrows indicate the sites of fusion of the sry β or δ and E. coliβ -galactosidase open reading frames in the pSBL1 and pSDL1 reporter genes. The position of each finger (numbered from 1 to 7) is represented by an open box; the shadowed area represents the putative His-His α -helical loop according to Berg (1988). Revision of the structure of the sry β gene intron (Vincent et al. 1985), as determined by S1 mapping and cDNA sequencing (data not shown) added 5 residues to the sry β protein sequence: Gly, Cys, Ile, Vai, Pro after the Glu residue position 25. Corrections in the sry δ protein sequence include additional residues: Pro at position 62, Asp at position 162 and Vai (in place of Ser)-Leu-Lys at the C-terminal end, and a.a replacements: Ser → Met position 273, Phe → Val position 394, Thr → Asn position 414.

Fig. 1.

Conservation and divergence within the sry β compared to the sry δ protein. The representation is adapted from that chosen in Treier et al. (1989). Conserved residues are indicated by black lines. Deletions/insertions, as defined in the text, are represented by triangles facing downwards and upwards, respectively. The open, and small and large black triangles correspond to 21, 1 and 5 bp insertions, respectively. The following exchanges were considered as conservative ones: L⇌ V⇌ E⇌ D; S⇌ T; Y⇌ F; N⇌ Q. The position of the sry β intron is marked. Open arrows indicate the sites of fusion of the sry β or δ and E. coliβ -galactosidase open reading frames in the pSBL1 and pSDL1 reporter genes. The position of each finger (numbered from 1 to 7) is represented by an open box; the shadowed area represents the putative His-His α -helical loop according to Berg (1988). Revision of the structure of the sry β gene intron (Vincent et al. 1985), as determined by S1 mapping and cDNA sequencing (data not shown) added 5 residues to the sry β protein sequence: Gly, Cys, Ile, Vai, Pro after the Glu residue position 25. Corrections in the sry δ protein sequence include additional residues: Pro at position 62, Asp at position 162 and Vai (in place of Ser)-Leu-Lys at the C-terminal end, and a.a replacements: Ser → Met position 273, Phe → Val position 394, Thr → Asn position 414.

The sequence alignment of sry β and sry δ proteins (Fig. 1) requires only two insertions longer than one amino-acid, one of which, 5 amino-acids long, is at the position of the sry β intron which has no counterpart in sry δ. It is readily apparent that the highest degree of conservation between sry β and sry δ is found in two separate regions: (i) the N-terminal part of the protein between amino-acids 8 and 88, showing 44 % homology (ii) the finger domain, amino-acids 172 to 337, encompassing 6 adjacent fingers, which displays the highest degree of conservation. The degree of relatedness of individual fingers found at homologous positions in the two proteins is not, however, uniform. Fingers 1, 5 and 6 display 38%, 55% and 52% identical residues, respectively, when sry β and β are compared, while the values rise to 75 %, 65 %, and 75 %, respectively, for fingers 2, 3 and 4. The maximal homology is in the carboxy terminal part of the fingers which includes the His-His putative α helical loop (X1-L2-X3-X4-H5-X6-X7-X8-H9) and the linker that connects adjacent fingers (Berg, 1988). It must be noted that these linkers are not of the widespread form Thr-Gly-Glu-Lys-Pro (Schuh et al. 1986). Remarkably, sry δ displays 55 additional carboxy-terminal residues, including a 7th finger separated from the 6 others by 45 amino-acids (see Fig. 1).

The conserved N-terminal and finger regions of sry β and sry δ are separated by regions rich in negatively charged amino-acids. These acidic domains, sry β amino-acids 124 to 172 (negative charge –19) and sry δ amino-acids 90 to 183 (negative charge – 36) show very divergent primary sequences; there is an extra segment of 21 amino-acids in sry δ, which includes 14 consecutive Glu-Asp residues.

Anfi-sry β antibodies and transformant Drosophila SBL lines

To compare the expression of sry β and sry δ during embryogenesis we prepared antibodies directed against the sry β gene product. A slightly truncated form of the sry β protein was expressed in E. coli using the pSBAl plasmid (see Materials and methods). This polypeptide was purified and used to immunize rabbits. The specificity of the purified anti-sry β antibodies was demonstrated by immunoblotting assays against both bacterial and Drosophila protein extracts (Fig. 2B). In an E. coli lysate, the antibody recognizes only the pSBAl encoded sry β protein. In Drosophila nuclear extracts from wild-type embryos, the anti-sry β antibody reacts with a single protein of 39× 103Mr.

Fig. 2.

Specificity of the purified anti-sryβ and anti-sry δ antibodies. (A) Diagrammatic representation of the pSBLl and pSDLl reporter genes introduced into the Drosophila genome to give rise to SBL and SDL transformant lines, respectively. The open reading frames are drawn as boxes with the shaded and crippled areas representing the acidic and finger domains, respectively. (B) Total proteins extracted from induced E. coll cells carrying the pSBAl (pSBA) and pAR3039 (pAR) control plasmids, respectively, were immunoblotted with purified anti-sry β antibodies; 10 μg of proteins from nuclear extracts of 0 to 12 h old embryos from control ry506 (ry), SBL and SDL transformant lines were separated by electrophoresis on 10 % SDS PAGE, transferred to nitrocellulose and incubated with the purified antibodies: anti-sry β, (left); anti-sry δ (center); anti-β -galactosidase (right). The position of molecular weight markers is indicated by arrows on the left.

Fig. 2.

Specificity of the purified anti-sryβ and anti-sry δ antibodies. (A) Diagrammatic representation of the pSBLl and pSDLl reporter genes introduced into the Drosophila genome to give rise to SBL and SDL transformant lines, respectively. The open reading frames are drawn as boxes with the shaded and crippled areas representing the acidic and finger domains, respectively. (B) Total proteins extracted from induced E. coll cells carrying the pSBAl (pSBA) and pAR3039 (pAR) control plasmids, respectively, were immunoblotted with purified anti-sry β antibodies; 10 μg of proteins from nuclear extracts of 0 to 12 h old embryos from control ry506 (ry), SBL and SDL transformant lines were separated by electrophoresis on 10 % SDS PAGE, transferred to nitrocellulose and incubated with the purified antibodies: anti-sry β, (left); anti-sry δ (center); anti-β -galactosidase (right). The position of molecular weight markers is indicated by arrows on the left.

In order to distinguish between the maternal and zygotic contributions to the embryonic sry β protein, we made transformant Drosophila lines containing a sry β /β -galactosidase reporter gene. This reporter gene encodes a fusion protein that contains 340 (out of 356) NH2 terminal sry β residues, including the 6 zinc fingers (Fig. 2A, see Materials and methods). The SBL construct was introduced into the ry506Drosophila genome to establish transformant lines refered to below as SBL lines. Western blot analysis of embryo extracts using an anti-β -galactosidase monoclonal antibody revealed a nuclear protein specific to SBL lines with an apparent relative molecular mass of 155 × 103, the predicted size of the sry β/ β -galactosidase fusion protein (Fig. 2). The anti-sryβ antibody also recognises the 155 × 103Mr fusion protein; this strongly suggests that the 39 × 103Mr protein is sry β.

Comparison of results obtained with extracts prepared from ry506 control, SBL and SDL lines and reacted with either anti-sry β or anti-xry δ antibodies (see Payre et al. 1989 for description of the anti-sry δ antibodies and SDL transformant lines) confirmed that these antibodies are highly specific for sry β and sry δ, respectively, and do not show any detectable crossreactivity. The data show that, like sry δ, the sry β and β -galactosidase fusion proteins are nuclear proteins.

The sry β, sry δ and sry/ β -galactosidase fusion proteins were localised in embryos by an immunostaining procedure (Caroll and Scott, 1985) using anti-sryβ, anti-sry δ (Fig. 3A), and anti-β -galactosidase (Fig. 3B) antibodies, respectively.

Fig. 3.

Compared expression of sry β and sry δ in embryos. Whole-mount preparations of fixed embryos were immunostained according to Carroll and Scott (1985). Embryos are oriented so that anterior is to the left and ventral is down. (A) ry506 embryos stained with anti-sry β (left), or anti-sry δ (right) antibodies. (B) embryos from ♀ ry506 × ♂ SBL or ♂ SDL and reciprocical crosses, as indicated at the top, stained with anti-β -galactosidase antibody. Similar stages are shown in panels A and B. (a) Embryos undergoing the eight nuclear cleavage; (b) blastoderm stage (interphase of the 14th nuclear division cycle) embryos. The stronger sry β staining, compared to sry δ (panel A), is contributed by both maternal and zygotic synthesis of the sry β protein present at that stage (see panel B); (c) embryos during germ band elongation (stage 8) and (d) during (stage 12, panel B) or after (stage 13, panel A) germ band shortening show similar patterns of staining for sry β and sry δ. First zygotic expression of sry β / β -galactosidase is seen during germ band retraction (panel B); (e) embryos after organogenesis is almost completed (stage 17): maximal accumulation of the sry δ protein in the brain and ventral nerve cord occurring at that stage is not so distinct in the case of sry β.

Fig. 3.

Compared expression of sry β and sry δ in embryos. Whole-mount preparations of fixed embryos were immunostained according to Carroll and Scott (1985). Embryos are oriented so that anterior is to the left and ventral is down. (A) ry506 embryos stained with anti-sry β (left), or anti-sry δ (right) antibodies. (B) embryos from ♀ ry506 × ♂ SBL or ♂ SDL and reciprocical crosses, as indicated at the top, stained with anti-β -galactosidase antibody. Similar stages are shown in panels A and B. (a) Embryos undergoing the eight nuclear cleavage; (b) blastoderm stage (interphase of the 14th nuclear division cycle) embryos. The stronger sry β staining, compared to sry δ (panel A), is contributed by both maternal and zygotic synthesis of the sry β protein present at that stage (see panel B); (c) embryos during germ band elongation (stage 8) and (d) during (stage 12, panel B) or after (stage 13, panel A) germ band shortening show similar patterns of staining for sry β and sry δ. First zygotic expression of sry β / β -galactosidase is seen during germ band retraction (panel B); (e) embryos after organogenesis is almost completed (stage 17): maximal accumulation of the sry δ protein in the brain and ventral nerve cord occurring at that stage is not so distinct in the case of sry β.

Expression of the sry β protein during embryogenesis; comparison with sry δ expression

The sry β protein is present in early nuclear cleavage stage embryos. Staining of nuclei above the uniform ooplasmic staining seen prior to fertilisation (data not shown) is detectable as early as the four nuclei stage. During intravitelline divisions and migration of nuclei to the periphery of the embryo (up to nuclear division cycle 8, but prior to pole bud formation, stage 2 of Campos-Ortega and Hartenstein, 1985) all nuclei, including the precursors of pole cell nuclei are stained (Fig. 3A, a). At the blastoderm stage, the sry β protein is present in all somatic nuclei but is detectable only in residual trace amounts in pole cell nuclei (Fig. 3A, b). During germ band elongation (stage 8, Fig. 3A, c) and germ band shortening (stage 13, Fig. 3A, d), the spatial patterns of expression of the sry β and sry δ proteins are virtually indistinguishable (see Payre et al. 1989 for a detailed description of sry δ expression), although the peak of accumulation of the sry β protein markedly precedes that of sry δ (see Fig. 3A a,b and c). The predominant localisation of the sry δ protein in the sub and supraoesophagal ganglia and the ventral nerve chord after dorsal closure (stages 16 – 17) is not so distinct in the case of sry β (Fig. 3A, e, see also Fig. 3B).

Maternal and zygotic contribution to the embryonic sry β/ and sry δ / β -galactosidase proteins

A pattern similar to that of sry β was obtained for the sryβ / β -galactosidase protein in SBL homozygous embryos. The maternal and zygotic contributions to the embryonic sryβ / β -galactosidase were distinguished by comparing the time course of accumulation of the fusion protein in embryos from the crosses ♀ ry506 × SBL and ♀ SBL × ♂ ry506. In ♀ SBL × ♂ ry506 embryos, the sry β / β -galactosidase is already detectable during intravitelline nuclear divisions (Fig. 3B,a). In ♀ ry506 × SBL embryos, the zygotic protein encoded by the paternal transgene is detected, starting at the 13th nuclear cleavage (Fig. 3B, b). From this stage onto the end of embryogenesis, the patterns of zygotic and maternal plus zygotic accumulation of the sry β / β - galactosidase protein remain virtually identical (Fig. 3B, c-e). Comparison of sry β / β -gal and sry δ / β - gal expression confirmed that sry β is turned on earlier than sry δ in embryos since expression of the sry δ /lacZ gene in embryos from the ♀ ry506 × SDL was not detected before germ band extension (stage 9 – 10 embryos) (Fig. 3B, d and Payre et al. 1989). It also confirmed that, contrary to sry β, sry ζ predominantly accumulates in the central nervous system after dorsal closure of the embryo (Fig. 3B, e), and that this accumulation is due to the zygotic expression of the protein.

sry β and sry δ bind selectively to DNA in vitro

As the target sequences of sry β and sry δ are not known, we used the empirical method previously used by Desplan et al. (1985) to detect DNA sequence specific recognition. As a first step 5au3Al digested λ phage DNA was incubated with the purified sry proteins produced in E. coli. The protein-DNA complexes formed upon incubation were immuno-precipitated. The results obtained for sry β are shown in Fig. 4A. In the absence of protein, or with anti-sry-δ antibodies, no

Fig. 4.

Sequence specific interaction of the sry β and sry δ proteins with bacteriophage λ and plasmid pRLXSI DNA fragments. Immunoprecipitated DNA fragments (120 mM KC1 salt conditions, see Materials and methods) were separated on a 5 % denaturing polyacrylamide gel. (A) Sau3Al digested λ phage DNA; (B) Sau3Al digested pLRXSl plasmid DNA. Lanes a,a′: input DNA; b,b′ immunoprecipitated material in the absence of added sry protein in the assay; c,c′ immunoprecipitates in the presence of protein and tRNA without competitor DNA; lanes d,d′, e,e′ and f,f′ show immunoprecipitates obtained in the presence of 200 ng, 400 ng and 1 μ g, of calf thymus competitor DNA respectively; lane g, immunoprecipitate in the presence of sry β protein but with anti-sry δ antibody. The size (bp) of the pRLXSI Sou3Al fragments is indicated. The 66, 78, 88, 48, 126, 188 and 279 bp fragments are derived from the λ phage DNA. The 75 and 258bp fragments derive from the pTZ18R plasmid vector, as indicated by an asterisk. The open and black arrowheads point to fragments bound by either sry β or sry δ, but not by both.

Fig. 4.

Sequence specific interaction of the sry β and sry δ proteins with bacteriophage λ and plasmid pRLXSI DNA fragments. Immunoprecipitated DNA fragments (120 mM KC1 salt conditions, see Materials and methods) were separated on a 5 % denaturing polyacrylamide gel. (A) Sau3Al digested λ phage DNA; (B) Sau3Al digested pLRXSl plasmid DNA. Lanes a,a′: input DNA; b,b′ immunoprecipitated material in the absence of added sry protein in the assay; c,c′ immunoprecipitates in the presence of protein and tRNA without competitor DNA; lanes d,d′, e,e′ and f,f′ show immunoprecipitates obtained in the presence of 200 ng, 400 ng and 1 μ g, of calf thymus competitor DNA respectively; lane g, immunoprecipitate in the presence of sry β protein but with anti-sry δ antibody. The size (bp) of the pRLXSI Sou3Al fragments is indicated. The 66, 78, 88, 48, 126, 188 and 279 bp fragments are derived from the λ phage DNA. The 75 and 258bp fragments derive from the pTZ18R plasmid vector, as indicated by an asterisk. The open and black arrowheads point to fragments bound by either sry β or sry δ, but not by both.

DNA was detected in the precipitate (Fig. 4A, lane b and g). Without competitor DNA, many λ DNA fragments were bound nonselectively (lane c). With increasing concentrations of competitor DNA, the binding became selective with 9 out of 80 resolved DNA fragments specifically precipitated (lanes d – f). To be able to compare precisely the DNA-binding specificities of sry β and sry δ, we subcloned a λ phage fragment which contained 4 of the immunoprecipitable fragments (see Fig. 4B). Sau3Al digestion of DNA from the resulting plasmid (pRLXSl) yielded 14 resolvable fragments ranging from 46 to 341 bp in size (Fig. 4B lanes a and a′). In the absence of competitor DNA, 6 and 7 fragments were quantitatively immunoprecipitated by sry β and sry δ, respectively (lanes c and c′). Only 4 of these were common to both proteins, the others being specifically bound by one or the other. It must be noted that the 258 and 75 bp fragments derive from the plasmid vector. Increasing concentrations of competitor DNA in the binding assay drastically decreased the amount of all immunoprecipitated DNA fragments in the case of sry δ, while specific binding of sry β to the 258, 126 and 66 bp fragments was partly resistant (Fig. 4B, lanes d – f and d′ – f′).

In vivo binding patterns of sry β and sry δ on polytene chromosomes are specific of each protein

Nuclei of late third instar larvae salivary glands contain large polytene chromosomes allowing precise mapping of protein-DNA interaction sites (see, e.g., Sass, 1982; Kabisch and Bautz, 1983; Pirrotta et al. 1988; Zink and Paro, 1989). Since both sry β and sry δ are expressed in the nuclei of larval salivary glands (Payre et al. 1989 and data not shown), we looked for their potential binding sites on the polytene chromosomes, using an indirect immunostaining procedure (see Materials and methods).

When squashed fixed chromosomes from wild-type larvae were reacted with purified anti-sryβ or anti-sry δ antibodies, a discrete number of chromosomal sites of various intensities, but neither the nucleolus nor the chromocenter, were stained (Fig. 5A). The sry β protein was localised at about 60 specific sites (Fig. 5A,a), 54 of which were cytologically mapped (Table 1). A much larger number of sry δ binding sites, about 200, were revealed with anti-sry δ antibodies (Fig. 5A,b). In either case, the staining of any particular region appeared constant when chromosome spreads from separate larvae were compared. Observation of immunostained chromosomes after counterstaining with Giemsa revealed that the stained sites are mostly localised in interbands or at the edge of strong Bridges bands (Fig. 5B). Detailed comparison of the sry /land δ binding sites focused on the X chromosome and selected regions of the autosomes containing strong sry β binding sites (Fig. 5B and Table 1). This comparison showed that not only some weak but also some of the strong sry β binding sites do not correspond to sry δ binding sites (e.g. sites at 3B3-4, 21C8 and 21D3, Fig. 5B) while others could be common to both proteins.

Table 1.

Location of sry β binding sites on salivary gland polytene chromosomes of Oregon R larvae

Location of sry β binding sites on salivary gland polytene chromosomes of Oregon R larvae
Location of sry β binding sites on salivary gland polytene chromosomes of Oregon R larvae
Fig. 5.

sry β and sry δ proteins bind to different sets of chromosomal sites. Protein binding sites were visualised with the peroxidase substrate DAB (brown) and chromomeric bands with Giemsa stain (blue). (A) Complete set of wild-type chromosomes stained with anti-sry β (a) or anti-sry δ (b) antibodies. (B) Magnification of sections 1 to 7 of the X chromosome; c and d show the binding sites of sry β and sry δ, respectively; c′, d′ show the same chromosomes counterstained with Giemsa; e, f: tip of the chromosomal arm 2L stained with anti-sry β (e) or anti-sry δ (f) antibodies, and counterstained with Giemsa. (C) Tip of the X chromosome from SBL (g) and SDL (h) transformant lines immunostained with anti-β -galactosidase antibody and counterstained with Giemsa. Black arrows point to selected sry β binding sites (at positions 3B3–4, 21C8 and 21D3) that do not correspond to sry δ binding sites (open arrows).

Fig. 5.

sry β and sry δ proteins bind to different sets of chromosomal sites. Protein binding sites were visualised with the peroxidase substrate DAB (brown) and chromomeric bands with Giemsa stain (blue). (A) Complete set of wild-type chromosomes stained with anti-sry β (a) or anti-sry δ (b) antibodies. (B) Magnification of sections 1 to 7 of the X chromosome; c and d show the binding sites of sry β and sry δ, respectively; c′, d′ show the same chromosomes counterstained with Giemsa; e, f: tip of the chromosomal arm 2L stained with anti-sry β (e) or anti-sry δ (f) antibodies, and counterstained with Giemsa. (C) Tip of the X chromosome from SBL (g) and SDL (h) transformant lines immunostained with anti-β -galactosidase antibody and counterstained with Giemsa. Black arrows point to selected sry β binding sites (at positions 3B3–4, 21C8 and 21D3) that do not correspond to sry δ binding sites (open arrows).

Binding of sry β /lacZ and sry δ /lacZ fusion proteins on polytene chromosomes mimics that of the native proteins

In view of the marked differences between the sry β and sry δ patterns of chromosomal sites, it was of interest to look at the binding of the fusion proteins. We therefore immunostained polytene chromosomes from SBL and SDL homozygous larvae, (3 and 2 independant transformed lines, respectively) with a monoclonal anti-β-galactosidase antibody (Fig. 5C). The use of a single antibody for this comparison rules out the possibility that a cross-reaction between anti-sryβ or anti-sry δ antibodies and other chromosomal proteins contributed to the observed patterns.

The patterns of binding sites observed for the sryβ / β - galactosidase and sry δ / β -galactosidase proteins were similar to those of the wild-type sry β and sry δ proteins, respectively, as exemplified by a detailed view of the tip of the X chromosome (compare Fig. 5B,c′ and 5C,g; 5B,d′ and 5C,h). As a control, no staining of ryr56 chromosomes was seen (not shown). Detailed comparison of the sry β and sryβ/ β -galactosidase patterns showed that the number and position of sites were conserved, with, however in some cases, a decrease in staining intensity of the fusion protein sites.

Eucaryotic transcription factors can be broadly divided into two classes (Biggin and Tjian, 1989): one class is expressed in highly regulated spatial and temporal patterns correlating with gene-specific regulatory functions. The second class is expressed in a variety of cells and may regulate a wide range of genes. The Drosophila sry β, sry δ and sry h-1 genes encoding Cys2/ His2 finger proteins maternally provided to the embryo could belong to this second class (Vincent et al. 1988).

We have chosen the sry β and δ genes - so far the only identified set of two closely related Cys2/His2 finger protein genes in Drosophila - as a model for evolutionary studies on the structure and function of this class of DNA-binding proteins.

The presence of acidic domains in sry proteins supports the prediction that they are transcription factors, since a significant net negative charge is characteristic of the transcription activating region of various gene control proteins (Ptashne, 1988). The divergence of the primary sequence of the acidic domain between sry β and δ contrasts with the degree of conservation of the rest of the protein (Fig. 1). In a perhaps related vein, the transcription activating domain of the yeast Gal4 transcription activator can be replaced by a number of other amino-acid sequences, related only by their negative charge (Ma and Ptashne, 1987).

The predicted DNA-binding domain of the sry β and δ proteins comprises 6 contiguous fingers, an additional carboxy terminal 7th finger being found in sry δ. The degree of relatedness between individual fingers located at homologous positions in the two proteins is variable (Fig. 1). This suggests that all the fingers were not all subject to equal evolutionary constraints and may have taken on specific individual DNA-binding functions. Those that have been best conserved between sryβ and sry δ are the contiguous fingers 2,3 and 4 (Fig. 2). The conservation includes the amino-acids at positions 3 and 4 of the His–His loop, which generally correspond to highly variable positions (Bellefroid et al. 1989; Nietfeld et al. 1989) and have been proposed as being essential to specific contacts between DNA and fingers (Berg, 1988). Conservation of these 3 fingers may be responsible for the binding of the two proteins to some common DNA fragments in vitro (Fig. 4). Divergence of the other fingers may in turn be related to the differences in sry β and δ DNA recognition specificities in vitro and in vivo (Fig. 4, 5 and Table 1). Assuming that each Cys2/His2 finger may potentially bind ≈5 bp of DNA, (see discussion by Klug and Rhodes, 1987), the DNA target size of sry β and sry δ proteins could be up to ≈30 bp long. Even considering that possibly not all the fingers are required for specific DNA-binding (Vrana et al. 1988; Payre and Vincent, 1988) and not all base pairs of the recognized sequence are essential, it is statistically unlikely that bona fide binding sites of sry β or sry δ would be present within the 48 kbp of λ phage DNA. Binding of the sry β and sry δ proteins to several λ DNA fragments correlated with the sensitiveness of this binding to low concentrations of competitor DNA suggests that it reveals low affinity sites involving only limited specific contacts between DNA and the proteins.

The binding specificity in vivo might be influenced by protein modifications not occurring in E. coli or by interactions with other Drosophila factors which were lacking in our in vitro assay. We therefore compared the DNA-binding patterns of sryβ and sry δ in vivo, on polytene chromosomes. Around 60 sites (Table 1) were observed for sry β, while a much greater number of sites were reproducibly observed for sry δ (Fig. 5A). Major puffs are not stained, indicating that neither sry (3 or sry δ proteins are general RNA polymerase II associated factors (Kabisch and Bautz, 1983). Cytogenetic mapping of sry β (and part of sry δ, not shown) binding sites did not show up any comprehensive gene family, partly due to the absence of precisely mapped genes in the regions corresponding to mapped sites (Merriam, 1988). The number of binding sites reported for Polycomb and Zeste, two proteins that do not possess fingers and have completely unrelated structures, is similar to that of sry β (Zink and Paro, 1988; Pirrotta et al. 1988). A good correlation was noticed between the cytological positions of a dozen Polycomb binding sites and the position of genes with which Polycomb genetically interacts (Zink and Paro, 1988). We have no direct evidence that the binding of sry proteins at all the sites seen on chromosomes is functional (Berg and Von Hippel, 1988). Nevertheless, the different patterns of sry β and sry δ suggest that they interact with specific, although probably partly overlapping, sets of genes. The marked difference in their number of chromosomal binding sites was rather unexpected and could have been contributed to by the at least 5 fold difference in relative amounts of these two proteins in salivary gland nuclei (data not shown). For example, overproduction of zeste in salivary glands results in a vastly greater number of bands binding the zeste protein than is observed in normal conditions (Pirrotta et al. 1988).

The patterns of binding sites of the sryβ/ β-galactosi-dase and sry δ/ β-galactosidase fusion proteins are similar to those of sry β and sry δ, respectively (Fig. 5B and C). It both confirms the sry β and sry δ DNA-binding specificities and shows that these specificities are maintained when the proteins are fused to the enzymatically active β-galactosidase. Decrease in staining intensity of some fusion protein binding sites could reflect a decreased accessibility of these sites to the larger proteins for steric hindrance reasons.. Alternatively, it could reflect site-specific competition effects between the wild-type and the chimeric proteins for binding to DNA. The chimeric protein approach used here should prove valuable for the identification of potential target genes for regulatory factors with no need for specific antibodies or the in vivo dissection of their DNA-binding domain. The present data indicate that the seventh carboxy-terminal sry δ finger (which is deleted in the fusion protein) does not account for the differences in the binding patterns of sry β and sry δ to polytene chromosomes. This can be brought together with data on in vitro mapping of TFIIIA DNA-binding domains, which showed that progressive deletions of its 4 carboxy-terminal fingers do not suppress its DNA-binding specificity and result in only a slight decrease in overall binding affinities of the mutant proteins (Vrana et al. 1988).

The presence of the sry β and δ proteins in embryonic nuclei at the onset of zygotic transcription (Edgar and Schubiger, 1986) suggests that both proteins play a role in the activation of the zygotic genome. Like sry β, the sry β protein present in embryos originates from both maternal and zygotic expression of the gene (Fig. 3). Detection of the zygotic sry β/ β-galactosidase protein as early as the syncitial blastoderm stage correlates with the pattern of transcription from a transposed modified sry β gene (Vincent et al. 1986). It confirms that 459 bp of upstream DNA are sufficient for maternal and zygotic expression of sry β and that the embryonic protein is mainly of maternal origin; sryβ expression is detected, but at a much lower level, at every other period of the fly life cycle examined. (Vincent et al. 1985 and data not shown). The zygotic sryδ/ β-galactosidase protein is expressed starting at the germ band extension stage (stage 9–10 embryos) but, at later stages, predominates over the maternal product. The sry δ mRNA and protein are also present in various amounts in larvae, pupae and adults (Vincent et al. 1985; Payre et al. 1989). How the temporal difference between the maximal levels of expression of sry βand δ proteins during embryogenesis contributes to their functional specificity is not known. Several δ mutant alleles but no β mutations have been isolated from three independent screens based on zygotic lethality (Crozatier et al. in preparation). It is therefore possible that the low level of zygotic sry β expression is dispensable for survival. In the same way, it may be noticed that the only known phenotype of mutations in SuHw, another Cys2/His2 finger protein expressed in various tissues (Spana et al. 1988), is female sterility. In the near future, a combination of genetic and biochemical approaches should enable us to understand in details of the respective functions of sry β and sry δ during Drosophila development.

We are grateful to François Amalric and Michèle Crozatier for critically reviewing the manuscript, Françoise Lemeunier for suggestions and Jean Antoine Lepesant for his continuous interest. We wish to thank Edmond Barbey for photography and Joëlle Maurel for editorial assistance. This work was supported by CNRS, INSERM (grant n° 881019) and the Fondation pour la Recherche Médicale. F.P., S.N. and V.L. were supported by fellowships from Ligue Nationale Contre le Cancer and Ministère de la Recherche et de l’Enseignement Supérieur.

Aguadé
,
M.
(
1988
).
Nucleotide sequence comparison of the rp49 gene region between Drosophila subobscura and D. melanogaster
.
Mol. Biol. Evol
.
5
,
433
441
.
Bellefroid
,
E. J.
,
Lecocq
,
P. J.
,
Benhida
,
A.
,
Poncelet
,
D. A.
,
Belayew
,
A.
and
Martial
,
J. A.
(
1989
).
The human genome contains hundreds of genes coding for finger proteins of the krilppel type
.
DNA
8
,
377
387
.
Berg
,
J. M.
(
1988
).
Proposed structure for the zinc-binding domains from transcription factor IIIA and related proteins
.
Proc. natn. Acad. Sci. U.S.A
.
85
,
99
102
.
Berg
,
O. G.
and
Von Hippel
,
P. H.
(
1988
).
Selection of DNA binding sites by regulatory proteins
.
Trends Biochem. Sci
.
13
,
207
211
.
Biggin
,
M. D.
and
Than
,
R.
(
1989
).
Transcription factors and the control of Drosophila development
.
Trends Genet
.
5
,
377
383
.
Brown
,
R. S.
,
Sander
,
C.
and
Argos
,
P.
(
1985
).
The primary structure of transcription factor TFlllA has 12 consecutive repeats
.
FEBS Lett
.
186
,
271
274
.
Campos-Ortega
,
J. A.
and
Hartenstein
,
V.
(
1985
).
The Embryonic Development of Drosophila melanogaster
.
Springer Verlag
,
Berlin/New York
.
Carroll
,
S. B.
and
Scott
,
M. P.
(
1985
).
Localisation of the fushi tarazu protein during Drosophila embryogenesis
.
Cell
43
,
47
57
.
Daniels
,
D.
,
Schroeder
,
J.
,
Szybàlski
,
W.
,
Blattner
,
F.
,
Sanger
,
F.
,
Coulson
,
A.
,
Hong
,
G.
,
Hill
,
D.
and
Peterson
,
G.
(
1982
).
In LambdaW
(eds
Hendrix
,
R.
,
Roberts
,
J.
,
Stahl
,
F.
and
Weisber
,
R.
)
Cold Spring Harbor Laboratory
,
New York
.
Desplan
,
C.
,
Thies
,
J.
and
O’Farrell
,
P. H.
(
1985
).
The Drosophila developmental gene, engrailed, encodes a sequencespecific DNA binding activity
.
Nature
318
,
630
635
.
Dignam
,
J. D.
,
Lebovitz
,
R. M.
and
Roeder
,
R. G.
(
1983
).
Accurate initiation by RNA polymerase II in a soluble extract from isolated mammalian nuclei
.
Nucl. Acids Res
.
11
,
1475
1489
.
Edgar
,
B. A.
and
Schubiger
,
G.
(
1986
).
Parameters controlling transcriptional activation during early Drosophila development
.
Cell
44
,
871
877
.
Evans
,
R. M.
and
Hollenberg
,
S. M.
(
1988
).
Zinc fingers: Gilt by association
.
Cell
52
,
1
3
.
Hoey
,
T.
and
Levine
,
M.
(
1988
).
Divergent homeobox proteins recognise similar DNA sequences in Drosophila
.
Nature
332
,
858
861
.
Jowftt
,
T.
(
1986
).
Preparation of nucleic acids. In Drosophila, A Practical Approach
(
Roberts
,
D. B.
ed.)
IRL Press
, pp.
139
157
.
Kabisch
,
R.
and
Bautz
,
E. K. F.
(
1983
).
Differential distribution of RNA polymerase B and non histone chromosomal proteins in polytene chromosomes of Drosophila melanogaster
.
EMBO J
.
2
,
395
102
.
Klug
,
A.
and
Rhodes
,
D.
(
1987
).
‘Zinc fingers’: a novel protein motif for nucleic acid recognition
.
Trends Biochem. Sci
.
12
,
464
469
.
Knochel
,
W.
,
Poting
,
A.
,
Koster
,
M.
,
El-Baradi
,
T.
,
Nietfeld
,
W.
,
Bouwmeester
,
T.
and
Pieler
,
T.
(
1989
).
Evolutionary conserved modules associated with zinc fingers in Xenopus laevis
.
Proc. natn. Acad. Sci. U.S.A
.
86
,
6097
6100
.
Lindsley
,
D. L.
and
Grell
,
E. H.
(
1968
).
Genetic variations of Drosophila melanogaster
.
Carnegie Inst. Wash. Publ
.
627
.
Ma
,
J.
and
Ptashne
,
M.
(
1987
).
A new class of yeast transcriptional activators
.
Cell
51
,
113
119
.
Merriam
,
J.
(
1988
).
The Drosophila clone list by chromosome location
.
D.I.S
.
67
,
11
135
.
Miller
,
J.
,
Mclachlan
,
A. D.
and
Klug
,
A.
(
1985
).
Repetitive zinc-binding domains in the protein transcription factor IIIA from Xenopus oocytes
.
EMBO J
.
4
,
1609
1614
.
Mitchell
,
P. J.
and
Than
,
R.
(
1989
).
Transcriptional regulation in mammalian cells by sequence specific DNA binding proteins
.
Science
245
,
371
378
.
Nagai
,
K.
,
Nakaseko
,
Y.
,
Nasmyth
,
K.
and
Rhodes
,
D.
(
1988
).
Zinc-finger motifs expressed in E. Coli and folded in vitro direct specific binding to DNA
.
Nature
332
,
284
286
.
Nietfeld
,
W.
,
El-Baradi
,
T.
,
Mentzel
,
H.
,
Pieler
,
T.
,
Koster
,
M.
,
Poting
,
A.
and
Knochel
,
W.
(
1989
).
Second-order repeats in Xenopus laevis finger proteins
.
J. molec. Biol
.
208
,
639
659
.
Payre
,
F.
and
Vincent
,
A.
(
1988
).
Finger proteins and DNA-specific recognition: distinct patterns of conserved amino acids suggest different evolutionary modes
.
FEBS Lett
.
234
,
245
250
.
Payre
,
F.
,
Yanicostas
,
C.
and
Vincent
,
A.
(
1989
).
Serendipity delta, a Drosophila zinc finger protein present in embryonic nuclei at the onset of zygotic gene transcription
.
Devi Biol
.
136
,
469
480
.
Pirrotta
,
V.
,
Bickel
,
B.
and
Mariani
,
C.
(
1988
).
Developmental expression of the Drosophila zeste gene and localisation of the zeste protein on polytene chromosomes
.
Genes and Development
2
,
1839
1850
.
Ptashne
,
M.
(
1988
).
How eucaryotic transcriptional activators work
.
Nature
335
,
683
689
.
Rosenberg
,
A. H.
,
Lade
,
B. N.
,
Chui
,
D. S.
,
Lin
,
S. W.
,
Dunn
,
J. J.
and
Studier
,
F. W.
(
1987
).
Vectors for selective expression of cloned DNAs by 77 RNA polymerase
.
Gene
56
,
125
135
.
Rubin
,
G. M.
and
Spradling
,
A. C.
(
1982
).
Genetic transformation of Drosophila with transposable element vectors
.
Science
218
,
348
353
.
Rubin
,
G.
and
Spradling
,
A.
(
1983
).
Vectors for P element mediated gene transfer in Drosophila
.
Nucl. Acids Res
.
11
,
6341
6351
.
Sass
,
H.
(
1982
).
RNA polymerase B in polytene chromosomes: immunofluorescent and autoradiographic analysis during stimulated and repressed DNA synthesis
.
Cell
28
,
269
278
.
Schuh
,
R.
,
Aicher
,
W.
,
Gaul
,
U.
,
Coté
,
S.
,
Preiss
,
A.
,
Maier
,
D.
,
Seifert
,
E.
,
Nauber
,
U.
,
Schroder
,
C.
,
Kemler
,
R.
and
Jáckle
,
H.
(
1986
).
A conserved family of nuclear proteins containing structural elements of the finger protein encoded by Krtippel, a Drosophila segmentation gene
.
Cell
47
,
1025
1032
.
Spana
,
C.
,
Harrison
,
A.
and
Corces
,
V. G.
(
1988
).
The Drosophila melanogaster supressor of Hairy-wing protein binds to specific sequences of the gypsy retrotransposon
.
Genes and Development
2
,
1414
1423
.
Tabor
,
S.
and
Richardson
,
C. C.
(
1985
).
A bacteriophage 77 RNA polymerase/promoter system for controlled exclusive expression of specific genes
.
Proc. natn. Acad. Sci. U.S.A
.
82
,
1074
1078
.
Treier
,
M.
,
Pfeifle
,
C.
and
Tautz
,
D.
(
1989
).
Comparison of the gap segmentation gene hunchback between Drosophila melanogaster and Drosophila virilis reveals novel modes of evolutionary change
.
EMBO J
.
8
,
1517
1525
.
Vincent
,
A.
(
1986
).
TFIIIA and homologous genes. The ‘finger’ proteins
.
Nucl. Acids Res
.
14
,
4385
4391
.
Vincent
,
A.
,
Colot
,
H. V.
and
Rosbash
,
M.
(
1985
).
Sequence and structure of the serendipity locus of Drosophila melanogaster. A densely transcribed region including a blastoderm-specific gene
.
J. molec. Biol
.
186
,
149
166
.
Vincent
,
A.
,
Colot
,
H. V.
and
Rosbash
,
M.
(
1986
).
Blastodermspecific and read-through transcription of the sry alpha gene transformed into the Drosophila genome
.
Devl Biol
.
118
,
480
487
.
Vincent
,
A.
,
Keuzlarova-Lepesant
,
J.
,
Segalat
,
L.
,
Yaniscostas
,
C.
and
Lepesant
,
J. A.
(
1988
).
sry h-1, a new Drosophila multifingered protein gene showing maternal and zygotic expression
.
Mol. Cell Biol
.
8
,
4459
4468
.
Vrana
,
K. E.
,
Churchill
,
M. E. A.
,
Tullius
,
T. D.
and
Brown
,
D. D.
(
1988
).
Mapping functional regions of transcription factor TFIIIA
.
Mol. Cell Biol:
8
,
1684
1696
.
Zink
,
B.
and
Paro
,
R.
(
1989
).
In vivo binding pattern of a transregulator of homeotic genes in Drosophila melanogaster
.
Nature
337
,
468
471
.