Cells need to control the location and timing of actomyosin-dependent force generation, and appear to do so in the first instance by regulating myosin filament self-assembly (Yumura and Fukui, 1985). The mechanism of the self-assembly is little understood. In vitro it is a true self-assembly, which requires a short domain at the C terminus of the myosin molecule. The availability of this domain appears suppressed by the folding of the molecule into a compact, looped state. In vitro, the rate at which these looped molecules unfold turns out to be a key determinant of filament number and filament length.

Long-tail, double-headed myosins self-assemble spontaneously in buffers of physiological ionic strength. The self assembly tethers the myosin head-pairs into ordered arrays, constraining sets of tens to several hundreds of them to join forces and pull along the same axis. It is not known if the proximity of other heads causes a tethered head-pair to function more efficiently. Certainly the time spent searching for a binding site on actin will be different for a head in a filament compared to one in solution, and it is possible that the heads in an array may restrict or guide each others’ diffusion in other ways. We would like to know more about the way self-assembly affects motor function, and to do this we need to understand the design principles of the myosin filament, particularly which of them are to do with the dynamics of assembly, and which are to do with optimising force generation. Specifically, we would like to know:

  1. Building units. Are filaments assembled from subfilaments or oligomers, or from monomers?

  2. Molecular packing. Is the relationship of one head to its partner optimised in some way? How much space does a myosin head-pair have, and how much freedom of movement radially and azimuthally? Are there cooperative structural effects?

  3. The choosing of partners. What are the dynamics of the assembly pathway? How fast can a filament be put together, and how fast can it be taken apart? What is the rate limiting step in both cases?

  4. Template-proteins. Which steps on the assembly pathway are perturbed by myosin-binding proteins?

  5. Molecular exchange. How fast do molecules exchange from the interior of the filament, relative to the ends? Can the rate of these processes also be affected by myosin-binding proteins? Does the presence of actin, for example, affect the dynamics of myosin self-assembly? How do other forces in the cell (mechanical, electrostatic, hydrostatic) affect exchange?

  6. Domain function. Which regions of the myosin molecule are important for assembly? Why does the myosin molecule have a long tail?

Here we discuss our own recent attempts to answer some of these questions, using smooth muscle and brush border myosins as model systems. On the basis of the data, we draw some tentative inferences about how ad hoc assembly and disassembly of myosin filaments might be achieved in motile cells.

Long-tail myosins equilibrate between two global conformations: a straight conformation, which self-assembles into filaments, and a looped or bent conformation which is incapable of self-assembly (Fig. 1). The assembly-incompetent folded conformation is populated to a different extent for different myosin isoforms, and according to the ionic conditions. We know that for the myosins we have studied, it is greatly stabilised by the binding of MgATP into the myosin active sites. The nucleotide is trapped in the active sites by the folding reaction (Cross et al. 1986, 1988).

Fig. 1.

Conformational species involved in myosin selfassembly. Straight-tail monomers equilibrate with filaments, and folded-tail monomers with straight-tail monomers. Myosin in solution in high salt was rapidly diluted into a large volume of 50 mM sodium formate, pH 7.0, to produce a final protein concentration of about 40 pg ml “. This sample was diluted with one volume of glycerol, vortexed, sprayed as a fine mist on to 5 mm squares of freshly-cleaved mica, dried under a vacuum of 2.10”5Torr for 30 min, and rotary shadowed at an angle of 6–8 degrees by slowly (30 s) evaporating 2 cm of 0.25 mm o.d. platinum wire wrapped around a tungsten carrier filament suspended about 10 cm from the specimen. Bar, 100 nm.

Fig. 1.

Conformational species involved in myosin selfassembly. Straight-tail monomers equilibrate with filaments, and folded-tail monomers with straight-tail monomers. Myosin in solution in high salt was rapidly diluted into a large volume of 50 mM sodium formate, pH 7.0, to produce a final protein concentration of about 40 pg ml “. This sample was diluted with one volume of glycerol, vortexed, sprayed as a fine mist on to 5 mm squares of freshly-cleaved mica, dried under a vacuum of 2.10”5Torr for 30 min, and rotary shadowed at an angle of 6–8 degrees by slowly (30 s) evaporating 2 cm of 0.25 mm o.d. platinum wire wrapped around a tungsten carrier filament suspended about 10 cm from the specimen. Bar, 100 nm.

Switching molecules from the looped into the straighttail state causes filament assembly. This can be effected (for smooth muscle and brush border myosins) by phosphorylation of the regulatory light chain, or (for molluscan myosins) by Ca2+-binding to the regulatory light chain, or (in both cases) by depletion of the MgATP in the medium. Self-assembly following depletion of the MgATP is rate-limited by the very slow (0.0002 s-1) loss of nucleotide from the folded molecules, thought to be due to transient excursions which they make into the straighttail state. On the basis of this assumption, it can be calculated that the straight-tail molecules comprise rather less than 1 % of the monomer population in equilibrium with filaments. Readdition of nucleotide to assembled filaments causes disassembly, as the folded monomer state is repopulated (Cross et al. 1986, 1988).

Smooth muscle and nonmuscle myosin filaments are thought to be composed only of myosin, no core or accessory proteins having been described. For smooth muscle native myosin thick filaments and for in vitro- assembled smooth muscle myosin, micrographs show filaments up to 5 μm long, composed of many hundreds of myosin molecules, with a 14.3 nm axial repeat in the head array, and/or a side-polar type of appearance (Sobieszek, 1972; Craig and Megerman, 1977; Hinssen et al. 1978; Cooke et al. 1989). In recent work, we have duplicated these findings and also obtained high resolution micrographs of the backbone structure of self-assembled smooth muscle myosin filaments, using a combination of improved specimen preparation and low dose electron microscopy. The images (Fig. 2) reveal that these filaments consist of sheets of straight, close-packed myosin molecules. The observation of close-packed, straight molecules assembled in a side polar arrangement with a 14.3 nm repeat defines the packing model shown in Fig. 3.

Fig. 2.

Fine structure of the tip of an in vitro self-assembled myosin filament. Myosin was polymerised by depleting the free MgATP in a 200μgml-l solution of folded-tail monomers in 150mM NaCl, 2mn sodium phosphate, 10mM imidazole, 1 mM MgCl2, 0.5 mM DTT, 0.5 mM EGTA, pH 7.3 (20°C). After the attainment of steady state, a 2μl drop of filaments was applied to a glow discharged carbon grid. After 5 s, the grid was drained by touching Whatman no. 1 filter paper to its edge, then rinsed with 6 drops of 1 % uranyl acetate, drained again and dried in air at room temperature. The grids were photographed at a nominal magnification of 43 000 under low dose conditions in a Philips CM12 electron microscope with a LaBs filament operated at 80 kV. Bar, 100 μm.

Fig. 2.

Fine structure of the tip of an in vitro self-assembled myosin filament. Myosin was polymerised by depleting the free MgATP in a 200μgml-l solution of folded-tail monomers in 150mM NaCl, 2mn sodium phosphate, 10mM imidazole, 1 mM MgCl2, 0.5 mM DTT, 0.5 mM EGTA, pH 7.3 (20°C). After the attainment of steady state, a 2μl drop of filaments was applied to a glow discharged carbon grid. After 5 s, the grid was drained by touching Whatman no. 1 filter paper to its edge, then rinsed with 6 drops of 1 % uranyl acetate, drained again and dried in air at room temperature. The grids were photographed at a nominal magnification of 43 000 under low dose conditions in a Philips CM12 electron microscope with a LaBs filament operated at 80 kV. Bar, 100 μm.

Fig. 3.

Scale diagram of filament packing. For clarity, only one head of each molecule is shown.

Fig. 3.

Scale diagram of filament packing. For clarity, only one head of each molecule is shown.

We have found that the rate of MgATP-induced disassembly depends on the filament geometry. Addition of ATP to very short myosin filaments (formed by rapid dilution of monomers into assembly conditions) causes them to disassemble completely within a few seconds. Longer filaments (formed by a slower reduction of the ionic strength at the same total myosin concentration (Sobieszek, 1972)), disassemble much more slowly, with a half time typically of about 30 s (Cross et al. 1986). This suggests that disassembly proceeds from the filament ends in both cases, since the only difference between the two populations is the number of filament ends. Electron microscopy of the disassembly reaction confirmed that this was so (Fig. 4). We concluded that the filament ends exchange molecules significantly faster than the interior of the filament, since otherwise the addition of MgATP would dissociate molecules from the interior of the filament and sever it.

Fig. 4.

Endwise disassembly of filaments, visualised by negative stain microscopy. Disassembly of a population of long filaments (top frame) was initiated by addition of an excess of MgATP. Samples were withdrawn from the reaction mixture at the intervals shown and applied to glow-discharged carbon grids. After 2–3 s the grid was rinsed with 1 % uranyl acetate, drained and the next sample taken. Conventional dose micrographs were recorded at a nominal magnification of 2500, which allowed entire grid squares to be observed. The areas shown were judged to be representative of the populations. Bar, 1 μm.

Fig. 4.

Endwise disassembly of filaments, visualised by negative stain microscopy. Disassembly of a population of long filaments (top frame) was initiated by addition of an excess of MgATP. Samples were withdrawn from the reaction mixture at the intervals shown and applied to glow-discharged carbon grids. After 2–3 s the grid was rinsed with 1 % uranyl acetate, drained and the next sample taken. Conventional dose micrographs were recorded at a nominal magnification of 2500, which allowed entire grid squares to be observed. The areas shown were judged to be representative of the populations. Bar, 1 μm.

The recovery phase which follows MgATP-induced disassembly yields further information about the assembly process. Titration of short myosin filaments with incremental amounts of MgATP produces incremental amounts of disassembly, and, as discussed above, subsequent reassembly. Fig. 5 shows that for substoichiometric amounts of MgATP, recovery produces a regain of essentially the original length distribution of the filaments. Recovery from superstoichiometric MgATP produces a dramatically different steady state, in which the filaments are much longer and fewer. We think the explanation for this is that reassembling molecules prefer to join on to the ends of existing filaments, rather than initiate new filaments. Substoichiometric MgATP shortens filaments but does not reduce their number, and reassembling molecules simply rejoin their ends. Superstoichiometric MgATP produces complete disassembly, so that reassembly must involve initiation of new filaments. Since this is unfavourable compared to growth, few filaments are initiated, and much growth occurs, resulting in a few, long filaments.

Fig. 5.

Pre-existing filaments as templates for assembly. Different amounts of MgATP were added to samples of the filament preparation shown in (A), causing various degrees of disassembly (not shown). As the MgATP was hydrolysed, reassembly occurred (Cross et al. 1986). The final length distribution of the filaments varied according to how much MgATP was added. For substoichiometric MgATP, disassembly was limited, and reassembly occurred on to the ends of existing filaments, producing a regrowth of the original population (B). For superstoichiometric MgATP, disassembly was complete, and reassembly required fresh nuclei to be formed. Nucleation is clearly unfavourable compared to elongation: rather few nuclei are formed, and much growth occurs, producing the population of long filaments shown in (C). Note that (B) and (C) are both stable situations. There was no detectable drift of the length distribution in either case. Bar, 1 μm.

Fig. 5.

Pre-existing filaments as templates for assembly. Different amounts of MgATP were added to samples of the filament preparation shown in (A), causing various degrees of disassembly (not shown). As the MgATP was hydrolysed, reassembly occurred (Cross et al. 1986). The final length distribution of the filaments varied according to how much MgATP was added. For substoichiometric MgATP, disassembly was limited, and reassembly occurred on to the ends of existing filaments, producing a regrowth of the original population (B). For superstoichiometric MgATP, disassembly was complete, and reassembly required fresh nuclei to be formed. Nucleation is clearly unfavourable compared to elongation: rather few nuclei are formed, and much growth occurs, producing the population of long filaments shown in (C). Note that (B) and (C) are both stable situations. There was no detectable drift of the length distribution in either case. Bar, 1 μm.

The repeating pattern of charge along the myosin tail has been suggested to specify the molecular overlaps for thick filament assembly (see the accompanying article by Atkinson and Stewart). Proteolysis data have gone some way towards testing this proposal, and have shown, contrary to expectation, that regions at or close to the C terminus of the myosin rod are crucial for self-assembly (Warrick and Spudich, 1987; Castellani et al. 1988). The precise location of the subdomain involved appears to vary according to the myosin isoform. In many myosins (but not mammalian skeletal muscle myosins) the C terminus of the tail carries an extension which is predicted to be nonhelical. This region typically contains phosphorylation sites, whose phosphorylation acts to prevent assembly.

Deletion mutagenesis in E. coli provides a powerful tool with which to dissect the domain function of the myosin tail at high resolution. Fig. 6 summarises recent results from our own laboratory in which progressively shorter versions of the brush border myosin II tail were expressed, purified, and assayed for self-assembly by airfuge ultracentrifugation and by electron microscopy. The results were striking: progressive deletion from the N terminus of the tail produced a progressive, but slight, increase in the solubility of the myosin tail. In contrast, deletion of the C terminus completely prevented self-assembly, even for otherwise full-length transcripts (Fig. 6).

Fig. 6.

The solubility of brush border myosin II rods, truncated by deletion mutagenesis. Mutant proteins were expressed in E. coli using a PIN expression vector. Fragments 2a, 3a and 2b were produced from cDNA clones isolated from a Igtll library, 2ab and 2AcΔ were constructed from the 2a DNA clone using restriction endonuclease sites within the clone. The sizes of the expressed fragments are shown as the number of peptide residues including those derived from the fusion between the myosin coding region and the E. coli Ipp gene in the expression vector .The solubility of expressed myosin rod fragments was tested by dialysing 100,id (0.5mgml-1) into 100mM NaCl, 25mM sodium phosphate, pH7.5, 0.5mM DTT and 2mn MgCl2 for 5 h. The recovered samples were weighed to determine final volume, then centrifuged at 100 000 g for 20 min in a Beckman airfuge. A supernatant sample was taken for SDS-PAGE and the remainder of the supernatant removed. The pellet was resuspended in 100 μl 0.6 M NaCl, 25 mM sodium phosphate, pH 7.0, 10 mM Tris, pH 7.5, 0.5 mM sodium azide and 2 HIM DTT overnight. Pellet and supernatant samples were run on SDS-PAGE. The gel was stained with Coomassie Blue, destained and the percentage of myosin fragment in the supernatant fraction determined by densitometry of the gel bands.

Fig. 6.

The solubility of brush border myosin II rods, truncated by deletion mutagenesis. Mutant proteins were expressed in E. coli using a PIN expression vector. Fragments 2a, 3a and 2b were produced from cDNA clones isolated from a Igtll library, 2ab and 2AcΔ were constructed from the 2a DNA clone using restriction endonuclease sites within the clone. The sizes of the expressed fragments are shown as the number of peptide residues including those derived from the fusion between the myosin coding region and the E. coli Ipp gene in the expression vector .The solubility of expressed myosin rod fragments was tested by dialysing 100,id (0.5mgml-1) into 100mM NaCl, 25mM sodium phosphate, pH7.5, 0.5mM DTT and 2mn MgCl2 for 5 h. The recovered samples were weighed to determine final volume, then centrifuged at 100 000 g for 20 min in a Beckman airfuge. A supernatant sample was taken for SDS-PAGE and the remainder of the supernatant removed. The pellet was resuspended in 100 μl 0.6 M NaCl, 25 mM sodium phosphate, pH 7.0, 10 mM Tris, pH 7.5, 0.5 mM sodium azide and 2 HIM DTT overnight. Pellet and supernatant samples were run on SDS-PAGE. The gel was stained with Coomassie Blue, destained and the percentage of myosin fragment in the supernatant fraction determined by densitometry of the gel bands.

How does this C terminal domain confer the ability to self-assemble? The evidence suggests that interactions between the myosin tails in this region are stronger than in others. In Fig. 2 the N terminal part of the tail can be seen to flare away from the thick filament backbone a considerable distance, leaving the C terminal parts of the tails interacting and buried. In the absence of other evidence, we imagine that the C termini interact directly with one another, and that this strong interaction is likely to be important for an initial recognition between myosin molecules which serves to orient them at an early stage of their binding to a growing filament.

The loose coupling of assembly and contractile activity

In several systems, myosin self-assembly occurs coincidently with an increase in the myosin MgATPase activity. It has accordingly been unclear if filament self-assembly is required for high MgATPase activity and, by implication, for contraction. Recent experiments using monoclonal antibodies have enabled the two functions of assembly and the MgATPase to be dissected away from one another: thus molecules prevented by monoclonal antibodies from folding became locked into filaments, yet were still regulated by phosphorylation. Furthermore, molecules locked into the straight-tail monomer conformation had essentially the same MgATPase activity as those in filaments. It can be concluded that in these cases, i.e. brush border (Citi et al. 1989) and smooth muscle (Trybus et al. 1989), assembly is not necessary for enzyme activity, but is loosely coupled to it by virtue of having the same switch.

Concluding comments

The experiments discussed have revealed some of the component reactions of myosins II self-assembly, and some unexpected features of their dynamics. Key findings are that assembly depends critically on a narrow domain at the C terminus of the myosin tail, and that molecular exchange is at the filament ends, is rapid, and occurs in preference to the initiation of fresh filaments. A predicted consequence for ad hoc self-assembly is that monomers will be have to be supplied to the assembly reaction faster than they can add to the ends of existing filaments, if new filaments are to be formed. A second predicted property of the system is that it should be possible to stabilise myosin filaments by capping their ends, in much the same way that F-actin can be capped.

Regarding the myosin molecule itself, the packing diagram obtained offers a rationale for the observed importance of the C-terminal region in self assembly. This domain terminates in a nonhelical region whose function is unknown, but which can be suspected of an involvement in self assembly, perhaps in the initial binding of an attacking monomer. The subsequent stages by which such a molecule samples alternative orientations and settles into position are obscure. It may be possible to throw light on them using further deletion mutagenesis of the myosin tail. The greater part of the myosin tail appears simply to act as a spacer arm. The binding of this section of the tail to its neighbours is weaker than that of the C-terminal part to its neighbours, but this may be significant in providing for slippage at some stage of monomer addition, or in achieving the delicate balance between the mechanical integrity of the filament and the exchangeability of its component molecules. Again, expression mutagenesis should tell us the answer. The distribution of myosin heads produced by the self-assembly of wild-type myosin tails can reasonably be expected to have been functionally optimised by the forces of evolution. On the other hand recent work has shown that kinesin is functional when conjugated to a tail made of part of spectrin (Yang et al. 1990). Is the myosin tail really such an exquisite piece of engineering? Can we conceive of an artificially improved mounting for the myosin head?

Castellani
,
L.
,
Elliott
,
B. W.
and
Cohen
,
C.
(
1988
).
Phosphorylatable serine residues are located in a non-helical tailpiece of a catch muscle myosin
.
J. Mus. Res. Cell Mot.
9
,
533
540
.
Citi
,
S.
,
Cross
,
R. A.
,
Bagshaw
,
C. R.
and
Kendrick-Jones
,
J.
(
1989
).
Parallel modulation of brush border myosin conformation and enzyme activity induced by monoclonal antibodies
.
J. Cell Biol.
109
,
549
556
.
Cooke
,
P. H.
,
Fay
,
F. S.
and
Craig
,
R.
(
1989
).
Myosin filaments isolated from skinned amphibian smooth muscle cells are side polar
.
J. Mus. Res. Cell Mot.
10
,
206
220
.
Craig
,
R.
and
Megerman
,
J.
(
1977
).
Assembly of smooth muscle myosin into side-polar filaments
.
J. Cell Biol.
75
,
990
996
.
Cross
,
R. A.
(
1988
).
What is 10S myosin for? J. Mus. Res. Cell Mot.
9
,
108
110
.
Cross
,
R. A.
,
Cross
,
K. E.
and
Sobieszek
,
A.
(
1986
).
ATP-linked monomer-polymer equilibrium of smooth muscle myosin: the free folded monomer traps ADP.Pi
.
EMBO J
.
5
,
2637
2641
.
Hinnsen
,
H.
,
DiHaese
,
J.
,
Small
,
J. V.
and
Sobieszek
,
A.
(
1978
).
Mode of filament assembly of myosins from muscle and nonmuscle cells
.
J. Ultrastruc. Res.
64
,
282
—302.
Sobieszek
,
A.
(
1972
).
Crossbridges on synthetic smooth muscle myosin filaments
.
J. molec. Biol.
70
,
741
—744.
Trybus
,
K. M.
(
1989
).
Filamentous smooth muscle myosin is regulated by phoshorylation
.
J. Cell Biol.
109
,
2887
2994
.
Warrick
,
H. M.
and
Spudich
,
J. A.
(
1987
).
Myosin structure and function in cell motility
.
A. Rev. Cell Biol. 3. 379-421
.
Yang
,
J. T.
,
Saxton
,
W. M.
,
Stewart
,
R. J.
,
Raff
,
E. C.
and
Goldstein
,
L. S. B.
(
1990
).
Evidence that the head of kinesin is sufficient for force generation and motility in vitro. Science
249
,
42
47
.
Yumura
,
S.
and
Fukui
,
Y.
(
1985
).
Reversible cyclic AMP-dependent change in distribution of myosin filaments in Dictyostelium. Nature
314
,
195
196
.