ABSTRACT
To provide a basis for studies of the expression of genes encoding the diverse kinds of intermediate-filament (IF) proteins during embryogenesis of Xenopus laevis we have isolated and characterized IF protein cDNA clones. Here we report the identification of two types of Xenopus vimentin, Viml and Vim4, with their complete amino acid sequences as deduced from the cloned cDNAs, both of which are expressed during early embryogenesis. In addition, we have obtained two further vimentin cDNAs (Vim2 and 3) which are sequence variants of closely related Viml. The high evolutionary conservation of the amino acid sequences (Viml: 458 residues; Mr ∼ 52800; Vim4: 463 residues; Mr∼ 53500) to avian and mam malian vimentin and, to a lesser degree, to desmin from the same and higher vertebrate species, is emphasized, including conserved oligopeptide motifs in their head domains. Using these cDNAs in RNA blot and ribonu clease protection assays of various embryonic stages, we observed a dramatic increase of vimentin RNA at stage 14, in agreement with immunocytochemical results ob tained with antibody VIM-3B4. The significance of very weak mRNA signals detected in earlier stages is dis cussed in relation to negative immunocytochemical re sults obtained in these stages. The first appearance of vimentin has been localized to a distinct mesenchymal cell layer underlying the neural plate or tube, respect ively. The results are discussed in relation to programs of de novo synthesis of other cytoskeletal proteins in amphibian and mammalian development.
Introduction
In embryonic development, cell differentiation and morphogenesis are correlated with - and probably dependent on - programs of expression of cell type specific proteins. This is especially evident for the diverse kinds of cytoskeletal proteins which contribute to the formation of major cell structures involved in cell and tissue shaping, in addition to other functions. Most cytoskeletal proteins, notably those forming the various kinds of filamentous structures, exist as multigene families whose members are differentially expressed during development in patterns specific for certain routes of cell and tissue differentiation. In amphibia, the best studied examples are the different actins (Sturgess et al. 1980; Vanderkerckhove et al. 1981; Mohun et al. 1984; Gurdon et al. 1985a) and the large family of proteins that form either cytoplasmic inter mediate-sized filaments (IFs) or the nuclear lamina (Franz et al. 1983; Benavente et al. 1985; Dawid et al. 1985; Godsave et al. 1986; Wylie et al. 1986).
The cytoplasmic IFs are formed by a highly complex group of polypeptides that are expressed in cell type characteristic patterns (for reviews see Anderton, 1981; Franke et al. 1982b; Lazarides, 1982). (1) Cytokeratins constitute a group of about 30 polypeptides that in cludes the approximately 20 polypeptides forming the typical IFs of epithelial cells (‘epithelial cytokeratins’; Franke et al. 1982b; Moll et al. 1982) and at least 10 different polypeptides characteristic of hair- and nail forming cells (‘trichocytic cytokeratins’; cf. Heid et al. 1988). (2) Typically, mesenchymally derived tissues contain IFs formed by vimentin (Franke et al. 1978) which, however, can also be coexpressed in some other cell types, including certain epithelia, with one of the other IF protein types (e.g. Franke et al. 1979, 1982b; Lazarides, 1982; Czernobilsky et al. 1985; and refer ences cited therein). (3) Desmin is an IF protein typically induced during myogenesis (for reviews see Lazarides, 1982). (4) The glial filament protein is a polypeptide synthesized typically, but not exclusively, in astrocytes. (5) The neurofilament polypeptides NF L, NF-M and NF-H are typically found in cells of neuronal differentiation (e.g. Liem et al. 1978; Anderton, 1981; Dahl & Bignami, 1985; Lasek et al. 1985).
Because of their abundance and cell type-restricted expression, the IF proteins are valuable candidates in studies of developmental changes of protein expression, in relation to cell differentiation and tissue formation. Various authors have used IF protein antibodies to identify the developmental appearance of the specific protein, by using immunocytochemistry alone or in combination with gel electrophoretic separations of cytoskeletal proteins. During mammalian development, cytokeratins 8 and 18, which form the first IFs, have been detected in mouse embryogenesis as early as the morula-blastocyst transition (Brûlet et al. 1980; Jackson et al. 1980, 1981; Franke et al. l982a,b) or even one or two cell generations earlier (Oshima et al. 1983; Duprey et al. 1985; Johnson et al. 1986; Chisholm & Houliston, 1987). Lehtonen et al. (1983a; see also Lehtonen, 1985, 1987) have reported that some cyto keratin already exists in the oocyte and the egg, although in a nonfibrillar, perhaps ‘soluble’ form (for divergent reports from other mammalian species see also Czernobilsky et al. 1985; Gall, LeGuen & Hun neau, 1988). In contrast, vimentin IFs have first been seen at day 8 of mouse embryogenesis, in primary mesenchymal cells (Franke et al. 1982a) and in some distal (parietal) endoderm cells that coexpress cyto keratins and vimentin (Franke et al. 1983; Lane et al. 1983; Lehtonen et al. 1983b). IFs containing desmin or the two neural types of IF proteins have not been detected in these early stages (e.g. Jackson et al. 1981; Raju et al. 1981; Schnitzer et al. 1981; Bignami et al. 1982; Bignami & Dahl, 1984; for avian embryogenesis see Tapscott et al. 1981). These observations suggest that in mammalian development early embryogenesis is dominated by the synthesis of an epithelial type IF cytoskeleton, in agreement with the polar architecture and the presence of typical desmosomes of ectodermal and endodermal cells.
In developmental research, amphibian embryogen esis, notably that of Xenopus laevis, is a ‘classical’ system for various reasons, not the least being the rapidity of early cleavages, blastula formation and gastrulation, which apparently is made possible by large maternal supplies of proteins and mRNAs. If one accepts the widely believed hypothesis that no tran scription of zygote genes takes place before the ‘mid blastula transition (MBT) point’ (for reviews see Kirschner et al. 1985; see, however, also Nakakura et al. 1987), one has to conclude that the early differentiation of epithelial cell layers, including the formation of cytokeratin IFs, that occurs before this point of devel opment, must use maternally supplied cytokeratins and cytokeratin mRNAs. Indeed, cytokeratins of the ‘sim ple epithelium type’, i.e. the amphibian equivalents to human cytokeratins 8, 18 and 19, and the corresponding mRNAs have been identified in oocytes, eggs and early embryos of Xenopus laevis (Franz et al. 1983; Gall et al. 1983; Godsave et al. 1984b; Wylie et al. 1985, 1986; Franz & Franke, 1986; Klymkowsky et al. 1987), whereas other kinds of cytokeratins are synthesized after gastrulation (for details see Fouquet et al. 1988).
With respect to vimentin synthesis, controversial immunocytochemical results have been reported. Franz et al. (1983) have not detected vimentin IFs in Xenopus oocytes of various stages but only in adjacent interstitial cells. In contrast, Godsave et al. (1984a) have reported reactions with vimentin antibodies in oocytes, eggs and throughout embryogenesis, suggesting that the protein is continually present but not specifically induced de novo at a certain developmental time. On the other hand, in this species, the synthesis of various myogenic proteins, including cardiac and skeletal muscle a-actins, has been elucidated in great detail, and it has been shown that their genes are postgastrulationally induced specifically in cells committed to a myogenic expression program (e.g. Mohun et al. 1984; Wilson et al. 1986; Mohun & Garrett, 1987; for a recent review see, Gurdon, 1987). To provide a basis for detailed examin ations of the embryogenic appearance, in space and time, of the specific IF proteins we have isolated cDNA clones of several IF proteins. In the present study, we describe cDNA clones for Xenopus vimentin which we have used for a study of their first expression, in combination with immunocytochemistry using a vimen tin-specific monoclonal antibody, VIM-3B4.
Materials and methods
Animals
Females of Xenopus laevis were kept as described (Krohne et al. 1981). After injection of female frogs with human chorion gonadotropin (Predalon, Organon, Oberschleissheim, FRG), eggs were stripped and fertilized in vitro. Embryos were incubated in 5 % DeBoers medium (100 % medium is: l10mm-NaCl, 1·2mm-KCl, 0·44mm-CaC(z, pH7·2; cf. Wolf & Hedrick, 1971) and staged according to Nieuwkoop & Faber (1967). After dejellying with 1·5 % cysteine-HCl (adjusted with NaOH to pH 7·8) in the same medium, the embryos were frozen in liquid nitrogen. For preparation of cytoskeletal material various embryonal stages, whole tad poles (stage 42), dissected material from adult animals and cultured X. laevis kidney epithelial cells (XLKE cells, line A6) were used (Franz et al. 1983).
Cytoskeletal preparations and gel electrophoresis Cytoskeletal proteins were prepared from cultured cells and oocytes as described (Franz et al. 1983), except that 3 mmPMSF and 1 mm-EGTA were included in the high-salt extrac tion buffer, and analysed by two-dimensional gel electrophor esis and immunoblotting. Vimentin antibodies were raised in guinea pigs (cf. Franke et al. 1979) against purified bovine lens vimentin commercially available from Progen, Biotechnics (Heidelberg, FRG). Monoclonal antibody PK Vl against human vimentin (Lehtonen et al. 1983b), which has been characterized on mammalian cells (Lehtonen et al. 1983b; Franke et al. 1984), was kindly provided by Dr I. Virtanen (University of Helsinki, Finland). Monoclonal antibody anti IFA reacting with most IF proteins (Pruss et al. 1981) was prepared in our laboratory from cultured supernatants of hybridoma cells purchased from American Type Culture Collection (Rockville, Maryland, USA). Monoclonal murine antibody VIM-3B4 (IgG2a) produced from mice immunized with purified bovine lens vimentin is available from Progen Biotechnics (Heidelberg, FRG) and Boehringer (Mannheim, FRG). In all species examined (human, bovine, rat, mouse, hamster, chicken, Xenopus laevis) it reacts specifically, in immunoblot tests and in binding assays using purified anti bodies, only with vimentin and does not cross-react with any other IF protein. The antibody is strongly positive on sections of frozen tissues as on sections of formalin-fixed, paraffin embedded tissue samples.
Soluble and cytoskeletal fractions of unfertilized eggs were prepared in two different ways. For the preparation of 100000g supernatants, 200 dejellied eggs were homogenized by pipetting up and down in 300 μl ‘5: 1’ medium (Callan & Lloyd, 1960) containing 3 mm-phenylmethylsulphonyl fluor ide (PMSF). After 10 min centrifugation at 3500 g the pellet was saved, and the resulting supernatant was centrifuged for 1 h at 100000 g to separate particulate material (‘high speed pellet’; HSP) from the corresponding supernatant fraction (HSS). At each step of preparation, aliquots of the fractions obtained were taken for SDS-polyacrylamide gel electro phoresis (SOS-PAGE).
For further fractionation, low speed pellets were extracted with buffers containing high salt and 1 % Triton X-100 (Franz et al. 1983). Alternatively, dejellied eggs were taken up directly in a 100 mm-morpholinoethane sulphonic acid buffer containing 0·8 m-NaCl and 1 % Triton X-100 (Herrmann & Wiehe, 1983), resuspended and centrifuged briefly at 800g or 10000g. Pellets and supernatants were processed for gel electrophoresis. All steps were carried out at 4°Con ice.
Samples were boiled in SOS-containing buffer and exam ined by SOS-PAGE (Laemmli, 1970) or by two-dimensional gel electrophoresis according to O‘Farrell (1975), using the modifications of Garrels (1979). Immunoblotting reactions on nitrocellulose sheets were visualized by 1251-labelled anti bodies (for refs see Achtstatter et al. 1986) or by the alkaline phosphatase method, following the technical manual of the supplier (Promega, Madison, WI, USA).
Preparation and in vitro translation of RNA
RNA was prepared as described and used for in vitro translation (Magin et al. 1983; Franz & Franke, 1986; Fouquet et al. 1988). Translation products were characterized by two dimensional gel electrophoresis, using cytoskeletal proteins from XLKE-A6 cells for coelectrophoresis.
RNA blot analysis
Poly(A)+ or total RNA was analysed by electrophoresis on formaldehyde/agarose gels (Davis et al. 1986) or on agarose gels after denaturation with glyoxal (for refs see Jorcano et al. 1984). After transfer to nitrocellulose paper, hybridization was carried out with 32P-DNA from clone pXenViml (for details see below) obtained by labelling of purified restriction fragments with [a--32P]dATP using the random .-_Erimed label ling kit (Boehringer Mannheim, FRG) or with P-anti-sense RNA obtained by transcription with T3 polymerase of BamHI-restricted pXenViml, yielding a probe representing the complete clone. For calibration, a commercially available RNA molecular weight marker set was used (Bethesda Research Laboratories, Gaithersburg, MD, USA).
Isolation and characterization of cDNA clones
For isolation of cDNA clones encoding Xenopus vimentin, we screened a cDNA library in }.gtlO which had been prepared with poly(A)+ RNA from stage-17 embryos (kindly provided by Dr D. A. Melton, Harvard University, Cambridge, MA, USA). As hybridization probe, a gel electrophoretically purified (SphI-Aflll) insert of the hamster vimentin cDNA clone, pVim 2(Quax-Jeuken et al. 1983; kindly provided by Dr W. Quax, University of Nijmegen, Netherlands) was used.
The start of this fragment corresponds, in the protein, to a position immediately in front of the a--helical rod domain (for nomenclature see Weber & Geisler, 1984; Steinert et al. 1985) and ends 61 nucleotides downstream of the stop codon. Under our conditions, randomly-primed labelled fragments with average lengths of 80-120 bp were obtained, resulting in a mixed pool of probes representing various regions of the insert. Hybridization conditions for Southern blot analysis were adopted from those used by Quax et al. (1984) with genomic X. laevisDNA, employing a pVim2-related, over lapping hamster vimentin clone as probe (pViml; Quax Jeuken et al. 1983). The insert size of the positive clones obtained was determined by a mini-preparation method (Lewis & Cowan, 1986), followed by Southern blot analysis of EcoRI-digested DNA with the pVim2 probe (see above). From 18 positive clones, 4 with very strong (pXenViml-4) and 4 clones with strong signals, all containing inserts larger than 1·8 kb, were chosen and used for large-scale preparation of DNA.
For isolation of cDNA clones encoding human vimentin, we screened a human testis cDNA library in }.gtll (Clontech, Palo Alto, CA, USA) essentially as described above, employ ing the hamster clone pVim2.
Subcloning and DNA sequencing
The EcoRI fragments of the clones obtained as described above were purified by gel electrophoresis on low melting point agarose gels. After phenol extraction of gel slices, the DNA fragments were inserted into the EcoRI site of the Bluescript Ml3+ plasmid (Stratagene, La Jolla, CA, USA). The polypeptides encoded by these subclones were character ized by the hybrid-selection-translation method (for details see Jorcano et al. 1984; Fouquet et al. 1988) and by in vitro transcription from these vectors as described by the manufac turer, followed by in vitro translation of the transcripts (see above). For each clone, all PstI fragments were subcloned into the Pstl site of the M13mp9 vector and sequenced by the dideoxynucleotide sequencing method (Sanger et al. 1977). Deduced amino acid sequences were used for computer-aided comparison with available IF sequences, using a special program of the German Cancer Research Center. Clones of interest were fully sequenced, using appropriate restriction enzyme fragments in Ml3 vectors aud the exonuclease III/Sl nuclease deletion method of Henikoff (1984). In addition, at least one strand was completely sequenced by the chemical degradation method (Maxam & Gilbert, 1977).
Ribonuclease protection assays
The procedure of Melton et al. (1984) was used with minor modifications. Sense RNA transcripts were synthesized with TI polymerase from pXenViml cloned into the Bluescript plasmid after linearizing with Clal. Uniformly labelled anti sense RNA was made from the same construct with T3 polymerase after cutting with Hinfl yielding a polynucleotide probe of 424 residues, including 69 nucleotides of polylinker, using [32P-a-]CTP or GTP (800Cimmol-1). A second probe, containing the region coding for the conserved TYRKLLE GEE domain (‘TYRKLEGE-probe’), was obtained by exo nuclease III deletion of pXenViml linearized by Kpnl and Clal. The Sl nuclease-treated, blunted and religated clone contained nucleotides 1-1300. After transcription with T3 RNA polymerase the Hinfl-restricted plasmid yielded a 32P labelled, ‘anti-sense’ RNA of 263 nucleotides, including 21 nucleotides transcribed from the plasmid. The RNA probes were purified bf urea-polyacrylamide gel electrophoresis and 1 x 10 cts min-(∼ lfmole) were hybridized overnight at 45 to 65°C with 5μg total RNA each or with lOμg tRNA in a solution containing 80 % formamide, 40 mm-Pipes (pH 6·4), 400mm-NaCl and 1 mm-EDTA. In an other series of exper iments, we used 50 % formamide instead of 80 % (Cho et al. 1988). The mixtures were digested with lO μg ml-1 RNase A and 20i.u.m1-1 RNase Tl (both obtained from Boehringer) at 30°C for 1 h. To test if this amount of probe was in excess of the RNA to be quantified, various amounts (approximately 1, 2·5, 5 and 12·5 μg) from stage-39 embryos were each pro tected with lx105 cts min-1 of probe, and the recovered RNA was quantified by liquid scintillation counting. A linear increase was obtained, indicating that the amount of probe used was sufficient to protect all vimentin transcripts present in 12·5 μg RNA of stage-39 embryos. Sense RNA transcripts were synthesized from the corresponding clones after lineariz ing at unique polylinker sites followed by transcription with the appropriate polymerase.
Expression of proteins in E. coli
The Xenopus vimentin clones pXenViml and pXenVim4 were ligated into the EcoRl site of the expression vector 17A/B (Magin et al. 1987) which was kindly provided by H. Bujard (Center for Molecular Biology, Heidelberg, FRG), and recombinant vimentin was isolated from inclusion bodies.
Immunoftuorescence microscopy
Cryostat sections through snap-frozen tissues or whole em bryos were processed for immunofluorescence microscopy as described by Jahn et al. (1987), using murine monoclonal antibodies VIM-3B4 to vimentin and lu-5 to cytokeratins (Franke et al. 1987) or guinea pig antibodies to vimentin (cf. Franz et al. 1983). Alternatively, ovarian tissue and embryos were fixed for 2 h in 2 % trichloroacetic acid (TCA) as described by Godsave et al. (1984a,b), frozen and cryostat sectioned. Sections were either air-dried or briefly (approx. 1 min) treated with 2 M- or 3M-urea (in PBS) in order to produce partial loosening of IF structures and ‘unmasking’ (Franke et al. 1983b; Godsave et al. 1984a,b). Further procedures were as described elsewhere (Jahn et al. 1987).
Results
(A) Identification of vimentin in Xenopus with specific antibodies
Monoclonal antibody VIM-3B4 to vimentin, which in immunoblot tests reacts specifically only with vimentin, detected characteristic fibrillar arrays in cultured epi thelial A6 cells of X. laevis (Fig. 1A). Upon colcemid treatment, these vimentin-positive filaments collapsed into the typical perinuclear aggregates (Fig. 1B,C) previously shown for vimentin IFs of various cultured cells of mammalian and avian origin (for refs see Bennett et al. 1978; Franke et al. 1978; Lazarides, 1982). This antibody was very specific and sensitive, as also shown by the fact that it allowed detection of the typical bundles of vimentin IFs present in the nucleated red blood cells (Fig. 10,E), a cell-type known to contain vimentin IFs in amphibia (Gambino et al. 1984). Essen tially identical results were obtained with our guinea pig antisera against bovine vimentin (data not shown).
The protein component responsible for this antibody reaction was identified as a cytoskeletal polypeptide of Mr- 55 000, as estimated from SOS-PAGE and immu noblot results using cytoskeletons of cultured XLKE A6 cells, stage-42 embryos (tadpoles) and erythrocytes from adult animals (Fig. 2A-C). In direct SOS-PAGE comparison, this polypeptide migrated slightly faster than vimentin from several mammalian cells (e.g. Fig. 2A), confirming the observation of Nelson & Traub (1982). On two-dimensional gel electrophoresis, the immunoreactive protein from Xenopus erythrocytes was resolved into a group of three isoelectric variants slightly more acidic than a-actin (Fig. 2D). A nearly identical pattern was obtained with high-salt-buffer and detergent-resistant residues from XLKE-A6 cells (Fig. 2F; at the loading shown here, except that the more basic spot was not split into two separate com ponents). However, in XLKE-A6 cytoskeletons, the immunoblot reaction did not fully coincide with the Coomassie blue staining (Fig. 2E), as was best revealed by staining immunoblots with alkaline phosphatase conjugated secondary antibodies followed by protein staining with Ponceau S. From the major, more basic Coomassie blue-stained spot only the left part was positive (designated Vin Fig. 2E,F) whereas the more acidic variant (X in Fig. 2E,F) was completely negative.
The more acidic, immunoreactive variant (V′ in Fig. 2D-F) was positioned between X and X′ (Fig. 2E) whereas spot X′ was negative with the vimentin anti body. These results were confirmed by immunoblots with guinea pig antiserum to vimentin, followed by 1251-protein A reaction and Ponceau S-protein staining. Reprobing these blots with antibody VIM-3B4 showed a superposition of both reaction sites (data not shown). Our results also show that in XLKE-A6 cytoskeletons two non-vimentin polypeptides with similar electro phoretic properties occur which might represent a yet unidentified IF protein, probably a type I cytokeratin (cf. Franz & Franke, 1986).
On two-dimensional coelectrophoresis of XLKE cytoskeletal proteins with hamster vimentin, Xenopus vimentin displayed a slightly higher electrophoretic mobility in SDS and a lower isoelectric point (inserts in Fig. 2E and F; note that the antibody VIM-3B4 recog nizes vimentin from both species equally well). More over, two-dimensional gel electrophoresis of partly degraded vimentin from both Xenopus and hamster further showed that this antibody reacted with the Mr∼38000 α-helical rod fragment (data not shown). This indicates that the epitope recognized by VIM-3B4 lies in a highly conserved region of the vimentins that is not present in desmin or other IF proteins.
(B) Characterization of cDNAs for Xenopus vimentin
Using a cDNA for hamster vimentin, pVim2 (Quax-Jeuken et al. 1983), we screened a library in λgt10 fromXenopus stage-17 embryonal RNA. From 18 positive plaques carried through three rounds of screening, twogroups of clones were distinguished: one groupcomprising11 clones with intense reactions and another group of 7 clones with less intense but significant signals. One clone (pXenViml) from the first group with an insert of approximately l-8kb was subclonedinto the Bluescript vector and used in hybrid selection experiments with total RNA from stage-18 Xenopus embryos. The clone selected a mRNA encoding a polypeptide indistinguishable from authentic XLKE-A6 Xenopus vimentin on two-dimensional gel electro-phoresis (Fig. 3A,B). Moreover, the polypeptide obtained from this clone by transcription and translation in vitro comigrated with the major Coomassie-blue stained spot of XLKE-A6 vimentin upon two-dimen sional gel electrophoresis (Fig. 3C,D). Unexpectedly, one of the four clones selected (pXenVim4) yielded a translation product that, on two-dimensional gel electrophoresis, was slightly more acidic (Fig. 3F) and slightly less mobile, compared to the pXenViml-coded product (Fig. 3F). This indicated that at least two different vimentins exist in Xenopus laevis that are both coexpressed in stage-17 embryos. Since these clones yielded nearly identical restriction maps with various restriction enzymes, Pstl fragments of these clones were subcloned into M13 vectors and sequenced. Partial sequence comparison with the hamster cDNA sequence confirmed that we had isolated X. laevis vimentin cDNA clones.
(C) Nucleotide sequences and deduced amino acid sequences
The two clones, pXenViml and pXenVim4, were fully sequenced. The cDNA sequence (1688bp) of pXen-Viml and the deduced amino acid sequence are shown in Fig. 4A. The first ATG codon is found in the sequence AACATGG known to be optimal for initiation by eukaryotic ribosomes (Kozak, 1986). The reading frame ends after 1374 nucleotides at a TGA codon, thus defining a polypeptide of 458 amino acids, including the initial methionine, corresponding to a total molecular weight of 52 812. This value is somewhat lower than that estimated from SDS-PAGE, as observed for all other vimentins sequenced thus far (Quax et al. 1983; Ferrari et al. 1986; Zehner et al. 1987). The sequence of a second clone, pXenVim2, was determined in parts. the 5’-end, this clone starts 30 bp more downstream than pXenViml, contains an additional AGC codon (42 codons after the ATG; Fig. 4A, upward triangle) and a conservative base change from T to C 18 bases down-stream from this AGC (see Fig. 4A, downward triangle). In this clone, the 3’-untranslated region is 58bp longer but identical with pXen Viml in the overlapping sequence. Otherwise, the two clones were identical for several hundred base pairs (data not shown). Another clone, pXenVim3, was found to be identical with pXenViml and pXenVim2 in the 3’-end up from the nucleotide marked by an arrowhead (Fig. 4A) but extended further downstream to a AATAAA polyadenylation signal, which was followed 14 nucleotides further by seven adenosine residues assumed to rep resent the begin of the poly(A)tract. As this clone was identical with pXen Viml in all nucleotides comparable, we have combined the two sequences to complete the vimentin sequence as shown in Fig. 4A.
In contrast, clone pXenVim4 differed considerably from pXenViml. Clone pXenVim4 contained the whole coding sequence, starting 11 nucleotides before the presumptive initiation codon. The amino acid sequence of Vim4 varied in a total at 31 positions (encircled in Fig. 4B) which were spread rather evenly throughout the molecule: 8 in the head, 16 in the rod, and 7 in the tail. Nearly all changes were conservative. In addition, vimentin Vim4 contained a 4-amino-acid insertion three residues before the carboxy terminus. Overall, this vimentin, Vim4, is 92 % identical to Viml at the amino acid level, but only 62 % homologous to Xenopus desmin (compare Herrmann et al. 1988). Furthermore, its sequence is identical with the recently published 74 amino acid sequence derived from the genomic clone XIFl (Sharpe, 1988), in contrast to our XenViml sequence which differs from XIFl in two positions. In the 3’-untranslated region pXenViml and pXenVim4 are nearly identical, except for a 9/18 nucleotide mismatch (overlined in Fig. 4B) and several single base substitutions.
Comparison of the deduced amino acid sequence of pXenViml with the sequences of vimentin of chicken (Zehner et al. 1987) and hamster (Quax et al. 1983) revealed that Xenopus vimentin is highly homologous to the avian and mammalian proteins (Fig. 5): 85 and 86 % of the amino acids of the α-helical rod domain of Xenopus vimentin Viml (Fig. 5) are identical to those of hamster and chicken vimentin, respectively, and most of the changes are conservative. The tail domain shows 73 and 65 % identity, compared to chicken and hamster, and the head 57 and 58 %, respectively. Within the α-helical rod, extended regions were ident ical in all three species, for example a stretch of 44 amino acids before the end of the helix. Moreover, the number of amino acids in the rod is absolutely con served within all three species. Compared to hamster vimentin, the head of the Xenopus vimentin Viml is shorter by 6 amino acids and the tail is shorter by one. In total, Xenopus vimentin Viml is 7 amino acids, and Vim4 two amino acids, smaller than hamster vimentin, in agreement with their slightly higher SOS-PAGE mobility.
(D) Conservation of sequences in the 3′ -untranslated region
It has been previously noted that the 3′ -untranslated regions of the hamster and chicken vimentin genes share a high degree of identical sequences, with a polyadenylation signal approximately 300 nucleotides after the stop codon (Quax et al. 1983). However, this type of homology was not found in the human vimentin sequence reported by Ferrari et al. (1986) in which a region of 57 nucleotides precedes a poly(A) tail without a classical polyadenylation site at the approximate distance. Since the 3′ -untranslated region of the Xenapus vimentin clones showed regions of sequence homology with the chicken and hamster genes (Fig. 6), including the distance between stop codon and poly-adenylation site, we screened a human cDNA library and sequenced several vimentin-positive clones. All of them (a representative, pHumViml, is shown in Fig. 6) differed in their 3′ -untranslated region from the sequence described by Ferrari et al. (1986) in that they were highly homologous to the other three species (Fig. 7). On closer comparison, several clusters of high homology became apparent in all species (boxes I-VII in Fig. 6). Remarkably, the putative poly(A)-addition signal, which lies within a consensus sequence (box VII) TRPyTICAATAAATCTIPyRGAAA, is much further downstream from the end of the human clone reported by Ferrari et al. (1986) who considered the AATAAA around the stop codon as the polyadenylation signal. Consensus box VI is also highly conserved and contains a second poly(A)-addition signal in the chicken sequence which, however, seems not to be used (Zehner & Paterson, 1983; cf. Capetanaki et al. 1983).
(E) Expression of vimentin in tissues and during embryogenesis
Using Northern blot analysis with a 5’-end-specific probe for Xenopus vimentin, we detected a single ∼ 2·1 kb band of vimentin mRN A in all tissues and cell cultures examined, although at very different intensities of reaction (Fig. 7A). This size is larger than that of the two cDNA clones (see above), probably due to the absence of most of the poly(A) sequence and also some untranslated 5’ sequence in the latter. The signals were very low in tissue samples from oesophagus (lane 1), skeletal muscle (lane 4) and ovary (lane 7), markedly higher in skin (lane 3), cardiac tissue (lane 5) and liver (lane 6), and most prominent in XLKE-A6 cell cultures (lane 2).
Northern blot analysis was also done with RNAs from various stages of Xenopus embryos using glyoxal for denaturation and an antisense riboprobe corre sponding to the complete pXenViml and very stringent washing conditions (72°C; Fig. 7B). Under these strin gency conditions, no significant reaction was obtained with RNAs from unfertilized eggs and early embryos (stages 6·5, 9 and 11; Fig. 7B, lanes 1-4). From stage 14 on (lane 5), a signal was found which increased in intensity up to stage 39 (lanes 5-8) and then remained constant in swimming tadpoles (lane 9). When these embryonic RNA samples were examined at somewhat lower stringency (65°C) using formaldehyde for RNA denaturation, very weak signals at the position of vimentin mRNA were detected in stage-6·5 to -11 embryos (Fig. 7C, lanes 2-4) and, on longer exposure, also in eggs (Fig. 7C, lanes 1 and l’). However, the intensities of the signals at the position of vimentin mRNA that were obtained with eggs and stage-6·5 to -11 embryos were similar to those seen at the positions of 28S rRNA present in the same gels (Fig. 7C, lanes 1-4 and l’, 2’). Therefore, we treated the blots extensively with RNase A, upon which only the reaction at 2·1 kb, i.e. the position of vimentin mRNA, was retained in mRNA samples from stages 6·5 to 11 (Fig. 7D, lanes 2-4).
For further clarification and quantification we also performed RNase protection experiments, using a 3′ specific vimentin probe as well as a probe representing nucleotides 1061 to 1302 of pXenViml, i.e. the nucleo tide sequence most highly conserved among the diverse IF protein genes (‘TYRKLEGE-probe’). At low and moderate stringency (45° and 60°C) both probes were specific for vimentin mRNA under the conditions used (Fig. 7E, lanes 10 and 12). Notably, in vitro synthesized mRNA for cytokeratin 1/8, which is known to be abundantly present in eggs and early embryos (Franz & Franke, 1986), was not protected at all (Fig. 7E, lanes 9 and 11). With both vimentin probes we observed protection by RNA of eggs (Fig. 7F, lanes 1, 3, 5, 7) and stage-18 embryos (lanes 2, 4, 6, 8). Compared to egg RNA, however, a more than hundredfold higher pro tection was found with stage 18 RNA, as estimated from determinations of the radioactivity of the pro tected bands. These results were confirmed and ex tended by two sets of experiments performed at higher stringency (65°C; 50 and 80 % formamide), which showed very weak but significant protection bands in oocytes, unfertilized eggs and embryos of stage-6·5 and -9 embryos (Fig. 7G, lanes 1-3 and Fig. 7H, lanes 1-4). These reactions, however, were more than hundredfold lower than those obtained with mRNAs from stage-18 (Fig. 7G, lane 4), and -28 (lane 5 in Fig. 7G,H) embryo. In addition, we noticed a conspicuous protected band of approximately 160 nucleotides (Fig. 7F, lane 4, arrow head) which probably presents the intercept of nucleo tides 1447 and 1608 of pXenVim4 (cf. Fig. 4) flanked by mismatch regions between Viml and Vim4.
(F) Localization of vimentin in ovaries and early embryos
Immunofluorescence microscopy on frozen sections of several tissues of adult Xenopus laevis showed, with monoclonal antibody Vim-3B4 and the vimentin specific guinea pig antibodies, positive reactions in cells of connective tissue, endothelium and erythrocytes, whereas skeletal and cardiac muscle as well as various smooth muscles were negative (Fig. 1D and data not shown). In ovarian tissue, the interstitial cells, including endothelial cells and erythrocytes, were strongly stained (Fig. 8A-D). In contrast, vimentin was not detected either in vitellogenic oocytes (Fig. 8A,B) or in previtellogenic oocytes (Fig. 8C,D). Essentially, the same results were obtained when the immunofluor escence reaction was performed on TCA-fixed ovaries (data not shown). Monoclonal antibody PK Vl which reacts with mammalian vimentin in immunoblots and in perimitotic filament configurations (cf. Lehtonen et al. 1983; Franke et al. 1984), was totally negative on Xenopus tissues (data not shown).
When embryonic stages were examined by immuno fluorescence microscopy, cells with significant vimentin immunostaining were not seen before gastrulation. In stage-14 (‘neural plate’) embryos, groups of cells posi tive for vimentin were detected that appeared to belong to two thin cell layers located below the neural plate (Fig. 9) whereas the ectoderm, the notochord, the neural plate and the presumptive somite cells were negative. At this point of development practically all cells of the embryo were positive for cytokeratins (Fig. 9B), including the vimentin-positive cells. In stage-18 (Fig. 10A) embryos, the vimentin-positive cells were still restricted to a mesenchymal cell layer underlying the neuroectoderm of the neural groove.and tube. All other tissues such as somites, notbchord, ectoderm, endoderm and the neural groove tissue were negative for vimentin but positive for cytokeratin (Fig. 10B).
(G) Screening of egg cell fractions for vimentin
In view of the obvious discrepancy between our negative findings and the positive reports of vimentin in oocytes and eggs (Godsave et al. 1984a), we also used immunoblotting to examine the possible existence of this protein in various protein fractions of unfertilized eggs. Specifically, we studied the ‘high speed super natant’ (HSS), in which some vimentin might exist in a soluble form (e.g. Blikstad & Lazarides, 1983; Soellner et al. 1985), and the corresponding pellet in which IF protein polymers would accumulate. When proteins of HSS .fractions were analysed directly (Fig. 11A, lane 2) or after concentration by precipitation with acetone (Fig. 11A, lane 1) no immunoreaction was detected with monoclonal VIM-3B4 antibody (Fig. llB, lanes 1 and 2) and with guinea pig antibodies to vimentin (data not shown). In controls, Xenopus vimentin from XLKE-A6 cells was readily detected (Fig. llB, lanes 7). The high-speed pellet, which was applied as a sample corresponding to the material from 100 eggs, was also negative for vimentin.
High-salt-buffer- and detergent-resistant residues of low speed pellets (Fig. 11A, lane 3) contained three major Coomassie blue-stained bands with Mr values of ∼66 000, 58000 and 43000 (see also Franz et al. 1983) but, at the position of Xenopus erythrocyte vimentin, no Coomassie blue-stained band was detected (data not shown). Similar results were obtained when unfertilized eggs were lysed directly in buffers containing high salt and detergent and pellets obtained at 800g (Fig. 11A, lane 5), at 10000g (lane 4) from the resulting super natant, and at 10 000 g without any precentrifugation (lane 6) were examined. Immunoblot analyses of these fractions (Fig. 11B, lanes 3-6) did not reveal any vimentin, even after extensive development of the colour reaction, resulting in unspecific staining of yolk proteins (Fig. 7B). The corresponding soluble fractions were also vimentin-negative (data not shown).
In parallel, we determined the sensitivity of this assay for vimentin using even recombinant Xenopus vimentin produced by E. coli after transformation with an expression vector carrying XenViml. As seen in Fig. 11C and D, vimentin can be diluted from 1 μg (Fig. 11C, lane 2, and D, lane 1) to 5 ng and still be detected with the colour reaction produced by the lg coupled alkaline phosphatase (Fig. 11D, lane 6). This shows that the egg cytoskeletons contain less than 0·25 ng vimentin per egg, if any. The corresponding value for HSS fractions was even lower (,:;_:_;Q-17 ng per egg). The guinea pig antibodies to vimentin also did not react with any protein of cytoskeletal fractions from unfertilized eggs, even after extremely prolonged ex posure (Fig. 11E,F).
To examine these negative results obtained with monoclonal antibody VIM-3B4 and with guinea pig antibodies to vimentin, we used two further monoclonal antibodies recommended for the detection of vimentin, i.e. antibodies PK Vl and anti-lFA. Fig. 11G-L show the immunoblot results obtained with cytoskeletal pro teins from oocytes (Fig. 11G,I,K; results with egg cytoskeletons are similar, not shown) in comparison with mixtures of oocytes and XLKE-A6 cytoskeletal proteins (Fig. 11H,1,L). Of the various cytoskeletal proteins present in such preparations (Fig. 11G and H show the Ponceau S-stained nitrocellulose-blotted poly peptides), antibody PK Vl reacted only weakly with authentic vimentin of XLKE-A6 cells (Fig. 11J, arrow) but this reaction was not specific as certain cytokeratins were also stained (Fig. 11J) and at prolonged exposure on oocyte polypeptides additional components were stained, including some with only slightly lower electro phoretic mobility than vimentin (Fig. 11I, arrowheads). Anti-IFA also reacted with vimentin of XLKE-A6 cytoskeletons as well as with many cytokeratins and some unidentified, cytoskeletal components, including cytokeratin 1/8 and a pair-spot component of about 60000 (Fig. 11K,L). Both antibodies do not react with an oocyte polypeptide located at the position of vimen tin. These results make clear that, in studies of Xenopus cells and tissues, both antibodies, PK Vl and anti-IFA, cannot be considered exclusive and sensitive probes for the unequivocal demonstration of Xenopus vimentin.
Discussion
The Xenopus laevis cDNA clones encoding various forms of vimentin should provide valuable tools for studies of cell and tissue differentiation during amphib ian development, particularly those dealing with the formation of mesenchymal and myogenic tissues. In addition, our results have allowed the unequivocal gel electrophoretic identification of the two major vimentin polypeptides of this species and a comparison of these proteins in taxonomically distant vertebrate species such as amphibia, birds and mammals. Clearly, all forms of Xenopus vimentin identified in this study are considerably more acidic (see also Franz et al. 1983) than the component tentatively identified as vimentin by Nelson & Traub (1982), which probably was one of the many cytokeratin polypeptides present in XLKE A6 cells.
The comparison of the amino acid sequences of the Xenopus vimentins with those of higher vertebrates indicates a remarkably high degree of sequence conser vation during vertebrate evolution. On the basis of the relatively low number of amino acid exchanges, Xeno pus vimentin appears equally distant from the chicken and the mammalian protein, both showing 79 % identical amino acid residues. Remarkably, the sequence homology is not confined to the a:-helical rod domain (86 % and 85 % identity between avian and hamster vimentin) but extends to both the head (58 % and 57 % identical residues) and the tail (65 % and 73 %). Our study also reveals the conservation, in all three species, of nucleotide sequence elements of considerable lengths (up to 32 residues) in the 3′- and, more restrictedly, the 5’ -untranslated regions of the vimentin mRNAs, suggesting that important regulatory func tions, in translation or transcription, are located in these regions (for previous comparisons of avian and mammalian vimentin see Quax et al. 1983; Zehner & Paterson, 1985; Ferrari et al. 1986; Zehner et al. 1987). The genes encoding vimentin have been reported to be single copy genes in other species (e.g. Quax et al. 1984, 1985a; Zehner & Paterson, 1985; Zehner et al. 1987; Ferrari et al. 1986). Without doubt, our sequence data show the existence of at least three different Xenopus laevis vimentin forms representing different genes, which is somewhat at variance with an earlier conclusion of Quax et al. (1984) made on the basis of Southern blot experiments probing Xenopus DNA with the hamster sequence. The sequence differences ob served in Xenopus between different cDNAs for the same kind of IF protein, as shown for vimentin (this study) and several cytokeratins (Hoffmann et al. 1985; Miyatani et al. 1986), are difficult to interpret. Prob ably, the three different forms of vimentin identified in the present study represent three of at least four vimentin alleles that are expected in a tetraploid species such asXenopus laevis (cf. Kobel & DuPasquier, 1986).
Using immunological reactions with the sensitive and vimentin-specific monoclonal antibody VIM-3B4 and guinea pig antisera to vimentin, we have not detected significant concentrations (;:.:::0-25 ng) of vimentin in oocytes of various stages, in eggs and in pregastrulation embryos (for reactivity of antibody VIM-3B4 with embryonic vimentin see Figs 2B, 9A and 10A). This is in agreement with our present and previous immunoflu orescence observations (Franz et al. 1983) but seems to be at variance with Godsave et al. (1984a) who reported positive immunocytochemical and immunoblot results for vimentin in oocytes, eggs and stage-10 embryos (see also Wylie et al. 1986). While this article was under review, a report by Tang et al. (1988) appeared in which positive vimentin immunoblot results have been reported for skeletal muscle tissue and oocytes using monoclonal antibodies PK Vl and anti-IFA, which is in contrast to our results with the same and other anti bodies. In contrast, we have found that PK Vl and anti IFA also react with several other cytoskeletal polypep tides of slightly lower electrophoretic mobility but different isoelectric points. Surprisingly, the polypep tide identified as vimentin by Tang et al. (1988) is slightly more basic than desmin, which is also in contrast to our two-dimensional gel electrophoretic separations (see also Herrmann et al. 1989). At present, we cannot resolve this controversy. We have to empha size, however, that our quantitative comparisons show that the maximal amount of vimentin that may possibly exist in oocytes and eggs (:e::;0-25 ng) is more than two orders of magnitude Jess than the total cytokeratin present (approx. lO0ng).
Our results with Northern blot and nuclease protec tion assays show that the amounts of vimentin mRNAs in Xenopus embryos increase drastically, i.e. approxi mately hundredfold, upon gastrulation, i.e. between stage 9 on the one hand and stage 14 on the other. The very weak but significant reaction of a band with the same electrophoretic mobility as vimentin mRNA in RNA samples from oocytes, eggs and pregastrulation embryos suggests the occurrence of trace amounts of vimentin mRNA in these cells, in basic agreement with Tang et al. (1988), although the relative postgastrulation increase of vimentin mRNA appears to be lower in the latter study. The biological significance of the finding of such low concentrations of vimentin mRNA in eggs (:e::;0-08pg per μgRNA, i.e. ∼0·3pg mRNA per cell) and in early embryonic development is difficult to assess, notably in relation to our findings of the pres ence of very little, if any, detectable protein. Similarly, trace levels of mRNAs in Xenopus eggs and early embryos have recently also been reported for the cell adhesion protein, N-CAM (Kintner & Melton, 1987), as well as for the acetylcholine receptor protein and cardiac a-actin, both usually considered to be muscle specific proteins (Baldwin et al. 1988).
The great increase of the synthesis of vimentin mRNA, and the observed restriction of vimentin to certain cell layers, at approximately 16 h after fertiliz ation, i.e. at the neural plate stage, indicates that the vimentin genes are among those that are transcribed with much greater intensity and frequency after gastru lation in cell-type-specific ways (Carrasco et al. 1984; Jonas et al. 1985; Dworkin-Rastl et al. 1986; Harvey et al. 1986; Kintner & Melton, 1987; Fouquet et al. 1988). Remarkably, this increase of vimentin is almost coinci dent with that of several major proteins typical of myogenesis (Mohun et al. 1984; Gurdon et al. 1985a,b; Wilson et al. 1986; Gurdon, 1987; Kay et al. 1987; Mohun & Garrett, 1987).
By immunofluorescence microscopy the first appear ance of vimentin in Xenopus embryos is not seen in all cells, but is detected only in certain mesodermal cells of layers subjacent to the neural groove and tube. Interestingly, most if not all cells of this mesodermal layer are positive for vimentin whereas the adjacent notochord and the presumptive somite tissue are nega tive. Apparently, these mesenchymal cells producing vimentin are also positive for cytokeratins, similar to other situations in which these two kinds of IFs have been found to coexist in certain embryonic and fetal cells of other vertebrate species (e.g. Lane et al. 1983; Lehtonen et al. 1983b; Czernobilsky et al. 1985; Erick son et al. 1987; Jahn et al. 1987). At present we cannot decide whether the cytokeratins found in these mes enchymal cells, as in all other cells of embryos of these stages, represent residual proteins synthesized in the oocyte or in pregastrulation stages, or whether they are actually synthesized at this stage in these specific cells. Clearly, simultaneous synthesis of vimentin and cytokeratins takes place in certain adult non-epithelial tissues of Xenopus laevis such as endothelia and certain smooth muscles (Godsave et al. 1986; Jahn et al. 1987).
ACKNOWLEDGEMENTS
We gratefully acknowledge the expert technical assistance of Monika Brettel and Ralf Zimbelmann. The work has been supported in parts by the Deutsche Forschungsgemeinschaft.