Expression of intermediate filament proteins during development of Xenopus laevis: II. Identification and molecular characterization of desmin

Of the various IF proteins that are differentially synthesized during embryogenesis, desmin has been shown to be typical for myogenesis, i.e. the formation of the cross-striated muscle tissue of myocardium and skeletal muscles and of the various kinds of smooth muscles (Cooke, 1976; Lazarides & Hubbard, 1976; Small & Sobieszek, 1977; Bennett et al. 1978, 1979; Lazarides et al. 1982). Desmin has also been found in the dispersed ‘myoid cells’ of the thymus (Henry, 1968; Hayward, 1972; Wekerle et al. 1975; Engel et al. 1977; Kao & Drachman, 1977; Drenckhahn et al. 1979; Dardenne et al. 1987; Kirchner et al. 1988) and the testis (Virtanen et al. 1986), in certain endothelial cells and pericytes (Fujimoto & Singer, 1986; Stamenkovic et al. 1986), stellate cells of liver of some species (Burt et al. 1986; Mijazaki et al. 1986), in cells of the interfollicular reticulum of lymph nodes and other lymphoid organs (Franke & Moll, 1987; Toccanier-Pelte et al. 1987) and in certain tumours, including myosarcomas as well as tumours not commonly regarded as myogenic (Osborn & Weber, 1983; Denk et al. 1983; Altmannsberger et al. 1985; Ellis et al. 1985; Gatter et al. 1986; Leader et al. 1987; Brown et al. 1987; Norton et al. 1987).

During myogenesis, the proportion of vimentin originally predominant in the muscle precursor cells appears to decrease gradually, whereas desmin increases, in some kinds of muscle to the extent that vimentin is no longer detectable (Bennett et al. 1979; Granger & Lazarides, 1979; Lazarides et al. 1982; Osborn et al. 1982). Clearly, vimentin and desmin can copolymerize and form heterotypic IFs in certain muscle tissues and cell culture lines (Steinert et al. 1981; Quinlan & Franke, 1982). However, desmin can also be coexpressed with cytokeratins such as in fetal myocardium of human and chicken, in myoid cells of bovine testis, and in some smooth muscle cells, notably in fetal tissues (Longo et al. 1987; Huitfeldt & Brandtzaeg, 1985; Brown et al. 1987; Jahn et al. 1987; van Muijen et al. 1987; Kasper et al. 1988; Kuruc & Franke, 1988).

During early mammalian and avian embryogenesis, IFs containing desmin have not been detected (e.g. Jackson et al. 1980, 1981; Tapscott et al. 1981; Bignami & Dahl, 1984; Erickson et al. 1987), and its advent appears to be correlated with the synthesis of other myogenic marker proteins (e.g. Fuseler et al. 1981; Lazarides et al. 1982; Whalen et al. 1982; Danto & Fishman, 1984; Fishman & Danto, 1984; Hayward & Schwartz, 1986; Mahdavi et al. 1986; Robert et al. 1986). For early amphibian embryogenesis, in which the expression pattern of IF proteins is different in that most, if not all, cells contain cytokeratins (Franz et al. 1983; Godsave et al. 1984, 1986; Wylie et al. 1986; Herrmann et al. 1989), it has been reported that several typical muscle proteins, including α-actins, are first synthesized after gastrulation and are specifically induced in the forming muscle tissue (Sturgess et al. 1980; Gurdon et al. 1984; Mohun et al. 1984; Wilson et al. 1986; Mohun & Garrett, 1987; for a recent review, see Gurdon, 1987). However, the myogenic IF protein, desmin, has not yet been included in studies of amphibian embryogenesis. Therefore, we have isolated and characterized a desmin cDNA clone from an embryonic library and have used it, in combination with desmin-specific monoclonal antibodies, to identify the time and location of desmin synthesis in amphibian development.

Maintenance of animals, cloning and sequencing procedures as well as RNA preparations and RNA blot experiments were performed as described in Herrmann et al. (1989). The radioactive probe for Northern blot analyses was constructed from a BawHI/Hmdlll-restricted, 670 nucleotides long 5 ′end-fragment of pXenDesl (containing 25 nucleotides of polylinker), using the random prime labelling kit from Boehringer (Mannheim, FRG). It was applied at l ·2 · 2×10⁶ctsmin ml ⁻¹ hybridization solution.

For preparation of cytoskeletal proteins from the muscular layer of Xenopus oesophagus, 20 μm thick cryostat sections of oesophagus from adult animals were microdissected and extracted with buffers containing 1 ·5 M-salt and 1 % Triton X100 (Herrmann et al. 1989). Stage-18 embryos were extracted accordingly, and aliquots of high-salt-resistant fractions (‘cytoskeletal proteins’) equivalent to 50 embryos were loaded in each isoelectric focusing separation in two-dimensional gel electrophoresis.

Immunoblotting and immunofluorescence microscopy were performed as described (Herrmann et al. 1989), using a monoclonal antibody specific for desmin (Debus et al. 1983; from Boehringer, Mannheim, FRG).

In the course of screening an embryonic cDNA library for Xenopus vimentin we isolated two clones with insert sizes of about 3 kb that gave identical restriction maps which, on the other hand, were completely different from those obtained with the Xenopus vimentin clones. When one of these clones was used in hybrid-selection experiments with total RNA from Xenopus stage-18 embryos, mRNA was isolated that gave rise, after translation in vitro and two-dimensional gel electrophoresis, to two polypeptides of M_r∼ 53 000 and ∼49000, respectively, the larger of which was more basic than the smaller one (Fig. 1A,B). Both radioactive polypeptides comigrated with the most basic component of two prominent series of spots of cytoskeletal proteins from the smooth muscle layer of Xenopus oesophagus (Fig. 1A,B) identified as desmin-related by their specific reaction with a monoclonal desmin antibody (Fig. 1C). A cytoskeletal fraction from stage-18 embryos (Fig. 1D) revealed upon immunoblotting with a monoclonal desmin antibody two spots which corresponded to the two products of the hybrid-selection experiments (arrows in Fig. 1E). These results indicated that the isolated clone encoded desmin, and therefore we designated the clone pXenDesl. The M_r∼ 49000 component is clearly a desmin-related polypeptide, but its chemical relationship to the M_T 53000 polypeptide requires further clarification.

In Northern blot analyses of heart muscle RNA this clone reacted with two mRNA bands of ∼ 3·0 and 1·9 kb (Fig. 2, lane 1). With skeletal muscle RNA we noticed the same two bands but, in addition, another band of ∼2·6kb (Fig. 2, lane 2). In RNA samples from ovary (lane 3) and cultured XLKE-A6 cells (lane 4) no signals were detected.

The sequence of the desmin clone, pXenDesl, is presented in Fig. 3. It contains 2995 bp and the second ATG is found within the consensus sequence ACCATGPu (Kozak, 1986). The 116 bp 5′-untranslated region is followed by an open reading frame of 1374 bp coding for 458 amino acids which corresponds to a calculated molecular weight of 52799, including the initial methionine that is probably lost post-translationally. This value is slightly lower than that given for hamster desmin (cf. Quax et al. 1985) which contains 468 amino acids, corresponding to a molecular weight of 53 308. The TAA stop codon introduces a 1505 bp 3′-untranslated region that contains three possible polyadenylation sites at positions 1754, 2114 and 2905. Whether these poly(A) addition sites correspond to the mRNA bands of 1·9, 2·6 and 3·0kb in RNA blot experiments of cardiac and skeletal muscle (Fig. 2) or whether there are further downstream poly-adenylation sites remains to be examined.

Comparison of the amino acid sequence deduced from pXenDesl with those of chicken and hamster desmin showed a close relationship (Fig. 4). The amino acids in the α-helical rod show 83 % identical positions (270 out of 326) in all three species, most of the differing positions being conservative exchanges. Remarkably, this homology extends to the tail region, which displays 77% (Xenopus/chick) and 85% (Xenopus/hamster) identical amino acids. In direct contrast, the head region is, apart from the highly conserved first 20 amino acids, more variable in size and in sequence (58 and 65 % identical amino acids with respect to hamster and chicken). Interestingly, however, the three possible phosphorylation sites of the protein kinase A type RXS reported for chicken desmin (cf. Geisler & Weber, 1988) are also present in Xenopus desmin (positions 40, 47 and 62; the second site is modified in the Xenopus sequence to KGAS, and to KGS in hamster desmin).

Sequence comparison of vimentin and desmin in the three species shows that the α-helical rod and the tail are very similar in the two IF proteins, whereas the head region seems to tolerate a higher degree of variability. Therefore, it is especially notable that in both vimentin and desmin the head contains shortly after the initial methionine, a highly conserved nonapeptide SSYRRXFGG (X being isoleucine or methion ine in vimentin, and threonine in desmin; Fig. 5). Following this consensus sequence, a PSFS sequence is conserved in desmin of Xenopus and hamster, and a similar motif is seen in the corresponding position of chicken desmin. In desmin, the number and the sequence of amino acids preceding this consensus sequence is also well conserved. Moreover, at position 10 of desmin, an arginine is found which introduces a possible phosphorylation site (RXS) for protein kinase A that is not present in vimentin. This conserved sequence motif is also seen as TSYRRTFGP, in the ‘c73 IF protein’ (Leonard et al. 1988), a recently detected rat IF protein expressed in a particular subset of mammalian neuronal cells (‘peripherin’; Portier et al. 1984) that is highly homologous to desmin in the a-helical rod domain but shows little relationship to desmin in the head domain.

Using RNA blot analysis (Northern blots) we detected desmin mRNAs first in stage-14 embryos (Fig. 6, lane 5) whereas in stage 11 no signal was observed (Fig. 6, lane 4). Remarkably, all three mRNA species observed in skeletal muscle were identified in all these embryonic stages (Fig. 6, lanes 5 —9) although their relative amounts varied considerably: in stage 28 the signal at ∼l ·9 kb was highest, compared to the ∼2 ·6 and ∼3 ·0kb bands (Fig. 6, lane 7). The intensity of the reaction at ∼3 ·0kb increased over that at ∼2-6kb in swimming tadpoles (stages 42; Fig. 6, lane 9).

By immunofluorescence microscopy on frozen sections of several tissues of adult Xenopus laevis with a monoclonal desmin-specific antibody, we observed positive reactions in smooth muscle tissue of the intestine and the oesophagus, in cardiac and skeletal muscle (where the fluorescence was concentrated in the Z-bands) as well as in cells of the endothelium (data not shown). When embryonic stages were examined by immunofluorescence microscopy on cryostat sections of frozen embryos, intense immunostaining was seen in early somite tissue of stages 14 (data not shown) and 18 (Fig. 7).

In parallel sections and double label experiments, the same somite regions were also positively stained with cytokeratin antibodies, including monoclonal antibody lu-5 (data not shown; see also figs 9 and 10 of Herrmann et al. 1989), indicative of the coexpression of cytokeratin and desmin IFs in the somite cells.

The cDNA clone pXenDesl encoding Xenopus laevis desmin extends our knowledge about this IF protein to amphibia and represents a valuable probe for studies of myogenesis and the regulation of expression of musclespecific cytoskeletal proteins. Comparison of the deduced amino acid sequence with the sequences of avian (chicken) and mammalian (hamster, pig) desmin shows that desmin has been well conserved during vertebrate evolution (more than 81 % identical residues in Xenopus and hamster desmin). Despite the considerable general sequence homology between desmin and vimentin (for other species see Geisler & Weber, 1981; Weber & Geisler, 1984; Quax et al. 1985), which in Xenopus amounts to 71 · 3% identical amino acid residues in the rod portion, it has to be stated that the difference between desmin and vimentin is as great in the amphibian (182 amino acid exchanges) as in the avian (182 different residues) and mammalian (176 exchanges) species. This indicates that the separation of these two IF proteins, which probably are derived from a common ancestral gene (Franke et al. 1978; Quax et al. 1983, 1985), has occurred long before the evolution of the anuran amphibia, in agreement with previous data of Quax et al. (1984). Particularly noteworthy is the maintenance of certain common sequence features in the rod and tail domains of both proteins in otherwise more divergent regions, including a SSYRRXFGG motif near the aminoterminus that may represent important functional sites or signals.

In alignments of the vimentin and desmin sequences of Xenopus, chicken and hamster, we have noted several evolutionarily conserved features that provide distinguishing characteristics for vimentin and desmin, respectively, (i) The head domains of both proteins carry five or more possible kinase A type phosphorylation sites which, however, are arranged in patterns differing in the specific proteins. It will have to be seen whether these sites are indeed involved in regulations of the assembly of the specific IFs formed as it has been proposed by Geisler & Weber (1988; for vimentin see also Inagaki et al. 1987; 1988). (ii) Besides the nonapeptide common to the heads of both vimentin and desmin, the number and the kind of amino acids preceding this nonapeptide are also conserved in desmin but not in the two major vimentins. (iii) In the highly conserved a- helical rod, we have noticed a divergent site in coil IB, which starts at position 196 of Xenopus vimentin Viml (LQSFFQDVDNA), which in desmin is replaced by LAAFFADVDAA, resulting in different surface domains on the <Y-helix (amides in vimentin, compared to hydrophobic residues in desmin), (iv) The carboxytermini differ in that the vimentins show a very acidic tip, DDFE, compared to HEVL in the desmins.

Surprisingly, we have found two polypeptides reactive with both our desmin probes, i.e. by the monoclonal antibody and by cDNA-hybrid selection of mRNAs, followed by translation in vitro. While the larger (A7_r ∼ 53 000) and slightly more basic component appears to correspond to the A7_r52800 polypeptide encoded by clone pXenDesl, the component with higher electrophoretic mobility, corresponding to M_r∼ 49000, might represent a different, perhaps truncated, form of desmin. A proteolytic derivative of the latter component, although not fully ruled out at present, appears less likely in view of our in vitro translation results. More direct analytical work is needed to answer this question.

At both the protein and the mRNA level, we have not found desmin in appreciable amounts in oocytes, eggs or in early embryonic stages, including blastulae. With both the mRNA and protein probes, desmin has first been detected in a postgastrulation stage, i.e. stage 14, the neural plate stage. Probably, with greater sensitivity it might be possible to date the actual onset of synthesis of desmin mRNAs somewhat earlier. Obviously, these data and the demonstrated enriched, if not exclusive, occurrence of desmin in somite tissue correlate with the advent, in space and time, of several other muscle-specific proteins (e.g. Sturgess et al. 1980; Gurdon et al. 1984, 1985a,b; Mohun et al. 1984; Gurdon & Fairman, 1986; Wilson et al. 1986; Kay et al. 1987; Mohun & Garrett, 1987). It will be interesting to find the regulatory elements that are responsible for the concerted expression of this ‘myogenic program’.

In mammalian and avian embryogenesis, vimentin appears in mesenchymal cells before, and in a wider range of cells than, desmin, and the latter IF protein has been detected only in myogenic cells that initially contain vimentin as their only IF protein (e.g. Lazarides et al. 1982; Bignami & Dahl, 1984; Bennett et al. 1979). In Xenopus embryogenesis, however, we have so far not detected vimentin in the somites or presumptive somite cells, although at the same time this IF protein is easily demonstrable in certain adjacent mesenchymal cell layers (Herrmann et al. 1989). Our finding that, in Xenopus, the somites and the presumptive somite cells contain cytokeratin IFs, like practically all early embryonic cells, suggests that the expression of desmin and the formation of desmin filaments can begin in cytokeratin-synthesizing cells and is compatible with ongoing cytokeratin synthesis. This observation may seem surprising on first sight, but is in agreement with recent reports of the coexistence of cytokeratin and desmin IFs in certain adult smooth muscle tissues of Xenopus (Jahn et al. 1987) and human (Brown et al. 1987; Jahn et al. 1987; Norton et al. 1987; Kasper et al. 1988) as well as in chicken and human embryonic myocardium (Van Muijen et al. 1987; Kuruc & Franke, 1988). Apparently, there exist regulatory factors that permit the simultaneous synthesis of certain IF proteins in one kind of tissue but lead to a mutually exclusive synthesis of the same IF proteins in other tissues.

We thank Monika Brettel and Ralf Zimbelmann for expert technical assistance. The work has been supported in parts by the Deutsche Forschungsgemeinschaft.