Evidence for the Essential Role of Myosin Subfragment-2 in Muscle Contraction: Functional Communication between Myosin Head and Subfragment-2

In 1971, Harrington et al. put forward a hypothesis, in which helix-coil transition in the hinge region of myosin subfragment-2 (S-2) contributes to muscle contraction. The helix-coil transition hypothesis has been, however, ignored by muscle investigators over many years. In 1992, we worked with him to examine the effect of polyclonal antibody to myosin subfragment-2 (anti-S-2 antibody), and found that the antibody eliminated Ca2+-activated isometric force generation of skinned vertebrate muscle fibers without affecting MgATPase activity. Further studies using the same antibody indicated functional communication between myosin head and myosin S-2, including regulation of binding strength between myosin head and actin filament during Ca2+-activated contraction in vertebrate muscle fibers. These findings indicate that the swinging lever arm hypothesis, in which muscle contraction results from active rotation of myosin head converter domain, is incomplete because it ignores of the role of myosin S-2. Much more experimental work is necessary to reach full understanding of muscle contraction mechanism at the molecular level.


Theory Structural basis of muscle contraction
In 1954, Hugh E Huxley and Jean Hanson made a monumental discovery that muscle contraction results from relative sliding between actin and myosin filaments, which in turn is produced by cyclic attachment and detachment between myosin heads extending from myosin filaments and corresponding sites in actin filaments [1]. As shown in Figure 1A, a myosin molecule (MW, 450,000) consists of two pear-shaped heads (myosin subfragment-1) and a rod of 156 nm long, and is split enzymatically into two parts: (1) the rod of 113 nm long (light meromyosin, LMM) and (2) the rest of myosin molecule including two heads and a rod of 43 nm long (heavy meromyosin, HMM). The HMM can be further split into two separate heads (subfragment-1, S1) and a rod (subfragment-2, S-2). In myosin filaments, LMM aggregates to form filament backbone, which is polarized in opposite directions across the filament central region (bare region, as illustrated in Figure  1B). The S-2 rod serves as a hinge between the S-1 heads and the filament backbone, so that the S-1 heads can swing away from the filament to interact with actin filaments. The axial separation between adjacent pairs of S-1 heads is 14.3 nm.
Actin filaments consists primarily of two helical strands of globular actin monomers (MW, 41,700), which are wound around each other with a pitch of 35.5 nm. The axial separation of actin monomers in actin filaments is 5.46 nm ( Figure 1C). In vertebrate muscle, actin filaments also contain two regulatory proteins, tropomyosin and troponin. In relaxed muscle, interaction between S-1 heads and actin filaments is inhibited sterically by tropomyosin, When Ca 2+ , released from the sarcoplasmic reticulum, binds with troponin; it removes the inhibitory effect of tropomyosin to start cyclic actin-myosin interaction producing sliding between actin and myosin filaments, i.e. muscle contraction. In skinned muscle fibers, in which the surface membrane is removed by treatment with glycerol, the fibers can be activated to contract by externally applied Ca 2+ . Figure 1D shows longitudinal arrangement of actin (thin) and myosin (thick) filaments in muscle. Actin filaments extend in either direction from Z-line to penetrate in between myosin filament arrays, which are located central in each sarcomere, i.e. the structure between two adjacent Z-lines to serve as structural and functional unit of muscle. In each sarcomere, the region containing only actin (thin) filaments is called the I-band, while the region containing both actin (thin) and myosin (thick) filaments is called the A-band. When a muscle is activated to contract, relative sliding between actin and myosin filaments, coupled with ATP hydrolysis, occurs in such a way that actin filaments are pulled into myosin filament arrays, while the length of actin and myosin filaments remains constant.

Attachment-detachment cycle between myosin heads and actin filaments
After the establishment of the sliding filament mechanism, the remaining problem is "what makes the filaments slide?". Both the actin binding sites and the ATPase activity are localized in S-1 heads, which will hereafter be called myosin heads. It is generally believed that myosin heads play a major role in muscle contraction coupled with ATP hydrolysis. Figure 2 illustrates an idea that occurred to Hugh Huxley as soon as he discovered the myofilament arrays in muscle electron microscopically [2]. As the longitudinal periodicity of myosin heads is different from that of myosin-binding sites in actin filaments, their interaction takes place asynchronously. As shown in the upper diagram, a myosin head (located in the left) extending from thick myosin filament first attaches to a myosin binding site in thin actin filament. The attached myosin head then changes its configuration to produce a unitary sliding between actin and myosin filaments; in such a way that actin filament is moved in between myosin filament arrays (middle diagram). The change in configuration of myosin head corresponds to myosin head power stroke to produce nitary sliding between actin and myosin filaments. After completion of power stroke, myosin heads detaches from actin filament, and returns to its original configuration (lower diagram). The return of myosin head to its original configuration corresponds to myosin head recovery stroke. Thus, myofilament sliding goes on by repeated asynchronous attachmentdetachment between myosin heads and actin filaments. Considering the dimension of myofilament-lattice, the amplitude of myosin head power and recovery strokes may not much exceed ~10 nm [3].
Muscle is a machine converting chemical energy of ATP hydrolysis into mechanical work. The only substrate for myosin head ATPase is MgATP [4].
For biochemical experiments on myosin head ATPase reaction steps, HMM or S-1 in solution is used. Their ATPase activity is very slow, but increases ~200 fold when actin filaments are added. This phenomenon is called actin-activation of myosin ATPase activity, and has been studied intensively by many investigators by quickly mixing HMM or S-1 solution with actin filament solution and measuring release of ATP hydrolysis product, Pi and ADP. It should be noted, however, that (1) HMM or S-1 concentration in the actomyosin solution is kept to be a few micromolar, compared to myosin head concentration in muscle fibers of 100-200 micromolar, and (2) that the 3D miofilment-lattice structure is completely lost in actomyosin solution. Nevertheless, the most plausible sequence of actomyosin ATPase reaction steps, known as the Lymn-Taylor scheme [5], has been presented. As shown in Figure 3, this scheme contains attachment of myosin heads to, and their detachment from, actin filament.   In this section, we will describe our early electron microscopic experiments using the gas environmental chamber [6], since the results obtained seem to be consistent with shortening of S-2 region predicted by Harrington. As illustrated diagrammatically in Figure 7, the gas environmental chamber (will hereafter be called EC) is a cylindrical compartment (diameter, 2 mm; height, 0.8 mm) with upper and lower windows to pass electron beam. Each window is covered with a thin carbon sealing film (thickness, 15-20 nm) held on a copper grid with nine apertures (diameter, 0.1 mm for each). The carbon sealing film can bear pressure difference up to 1 atmosphere. The biological specimen placed on the lower carbon film is kept wet by constantly circulating water vapor (pressure, 60-80 torr; temperature, 26-28°C) through the chamber. The EC contained an ATP containing microelectrode with its tip immersed in a thin layer of experimental solution, to apply ATP to the specimen ion theoretically. The EC was attached to a 200 kV transmission electron microscope (JEM 2000EX; JEOL). The material used was synthetic filament, in which rabbit skeletal muscle myosin was bund on the surface of thick paramyosin filaments obtained from molluscan catch muscle. The myosin-paramyosin core complex filaments (diameter, 50-200 nm; length, 10-30 µm) were very easy to handle. To position-mark individual myosin heads, colloidal gold particles (diameter, 15 nm; coated with protein A) were attached to myosin heads via site-directed monoclonal antibody to distal region of myosin head. Further details of the methods have been described elsewhere [8,9]. Figure 8 shows electron micrograph of myosin-manner. Chemical energy derived from ATP hydrolysis is stored in myosin heads in the form of M-ADP-Pi during the recovery stroke, to be used as mechanical energy producing power stroke after attachment to A. To repeat attachment-detachment cycle, power and recovery strokes should be the same in amplitude, and opposite in direction. Although extensive studies including chemical probe experiments and time-resolved X-ray diffraction experiments were hitherto made to record movement of myosin heads coupled with ATP hydrolysis, no definite results were obtained until 1997 when we first succeeded in recording ATP-induced myosin head movement in hydrated, living myosin filaments using the gas environmental chamber attached to an electron microscope [6].

Proposal of the helix-coil transition theory by Harrington
William F Harrington was a prominent polymer scientist, and got interested in the mechanism of muscle contraction. Reflecting his research carrier, he focused attention on the hinge region of myosin S-2, by which myosin heads can move towards actin filaments. To detect possible structural changes in the hinge region in myosin S-2, he used the enzyme probe method, in which the rate of splitting of S-2 hinge region by proteolytic enzyme, chymotrypsin was compared between skinned muscle fibers in relaxed, contracting and rigor states. He found that the S-2 hinge region was split rapidly by chymotrypsin in Ca 2+ -activated, contracting fibers, while the S-2 hinge region in relaxed and rigor fibers was not readily split by chymotrypsin ( Figure  5) [7], indicating some structural changes in the S-2 hinge region in Ca 2+ -activated fibers. Based on this and other experiments, Harrington proposed a contraction model shown in Figure 6, which resembles that in Figure 4, except that myosin head power and recovery strokes are replaced by helix-coil transition (or melting) of S-2 hinge region [7].   First, we recorded filament images with an imaging plate system (magnification, 10,000X) two times at intervals of several min, and found that the position of each gold particle did not change appreciably, indicating that each myosin head undergoes thermal motion around a definite equilibrium position, so that the time-averaged mean position of individual myosin heads remains almost unchanged, as predicted by Huxley [10]. Being encouraged with this finding, we recorded the same filament images two times, one before and the other after application of ATP. As shown in Figure 9, we found that individual gold particles, i.e. individual myosin heads, moved in response to ATP in the direction parallel to the filament axis. The average amplitude of ATP-induced myosin head movement was 20 nm, a value too large to be accounted for by the movement of myosin heads. This implies that, in the complex filaments, myosin molecules tend to attach to the paramyosin filaments parallel to its long axis at their LMM region, while it's HMM region can move more or less freely. If this assumption is correct, the large amplitude of ATP-induced myosin head movement is likely to result from the helix to random coil transition occurring at the S-2 region, constituting evidence that the helix to random coil transition actually takes place under some conditions. Since the experiments were performed in the absence of actin, the observed myosin head movement is regarded to be its recovery stroke, which is believed to be the same as power stroke in amplitude and opposite in direction.

Effect of anti-S-2 antibody on isometric force production, force-velocity relation, and MgATPase activity in skinned vertebrate muscle fibers
Although Harrington's theory was unique and interesting, most muscle investigators ignored it. Nevertheless, Harrington intended to demonstrate the important role of S-2 hinge region in muscle contraction, and prepared a polyclonal antibody to S-2 hinge region (anti-S-2 antibody), and showed that the antibody reduced Ca 2+ - Figure 6: Schematic diagram illustrating the helix to (random) coil transition cycle occurring in myosin-S-2 hinge region proposed by Harrington [9]. Helical and random coil structures of the S-2 hinge region are expressed as straight and zig-zag shapes, respectively [7]. activated isomeric force generation and slowed down Ca 2+ -activated myofibrillar shortening [11,12]. In 1992, he asked me to work together using his anti-S-2 antibody, and we made the experiments to be described later with unexpected results. At that time, he appeared to change his mind, and intended to emphasize possible communication between myosin head and subfragment-2 in regulating actin-myosin interaction producing muscle contraction, and his intension was achieved, at least in part, by the experiments described below. Figure 10 is a typical result showing the effect of anti-S-2 antibody on Ca 2+ -activated isometric force and Mg ATPase activity, measured by NADH fluorescence, in skinned vertebrate muscle fibers. In agreement with the report of Harringon [12], the antibody reduced isometric force in a time-dependent manner, while Mg ATPase activity remained unchanged even when isometric force was reduced to zero [13]. As shown in Figure 11, Muscle fiber stiffness, measured by applying small sinusoidal vibrations (amplitude, 0.1% of fiber slack length) decreased with decreasing isometric force, reaching zero when isometric force was reduced to zero. We also obtained the force-velocity relation of anti-S-2 treated muscle fibers by applying ramp decreases in force to isometrically contracting fibers and recording resulting fiber shortening ( Figure 12). To our surprise, the force-velocity curves thus obtained showed that, despite marked reduction of steady isometric force caused by the antibody, the maximum shortening velocity Vmax remained unchanged, as shown in Figure 13. It can be seen that the force-velocity curves, obtained at various levels of reduced isometric force ( Figure 13A) are shown to be identical when they are normalized with respect to the maximum steady forces ( Figure 13B). When author   informed the results described above to Harrington, he was extremely happy with them, and stated in his mail, "This is a triumph. We can now demonstrate the fact that S-2 region can influence on myosin head-actin interface!". To our great pity, however, he suddenly passed away before the publication of our work. Reflecting his contribution to the field of polymer science and his personality, a two day-long memorial symposium was held at Johns Hopkins University, and we were selected to be the first speaker by the request of Mrs. Harrington, who knew that his husband suddenly passed away with happy feeling in mind.

Evidence that myosin S-2 regulates binding strength between myosin heads and actin filaments
As soon as we received anti-S-2 antibody from Harrington, we examined the point of attachment of the antibody in myosin molecule by observing rotary shadowed electron micrographs of the antibody (IgG)-myosin complex, and found that the antibody attached not only to S-2 hinge region, but also uniformly to the whole S-2 region (Katayama, unpublished observation). Dr. Julian Davis, who worked with Harrington for many years, made similar observation as ours.
Therefore, Harrington's antibody did not bind specifically to S-2 hinge region, but bound uniformly to the whole S-2 region. We will consider the surprising effect of the antibody on this basis.
The effect of anti-S-2 antibody, which attaches uniformly over the whole S-2 region, can be summarized as follows [13]: (1) It can reduce Ca 2+ -activated isometric force to zero; (2) It can also reduce muscle fiber stiffness to zero; (3) Both force and stiffness decreases in parallel with each other; (4) Despite marked reduction of isometric force and muscle fiber stiffness, the maximum shortening velocity Vmax remains unchanged; (5) Despite marked reduction of isometric force and muscle fiber stiffness, MgATPase activity of muscle fiber does not change appreciably. Results (1), (2) and (3) can easily be explained as being due to decrease in the number of myosin heads producing isometric force, while result (5) indicates that myosin heads with the antibody attached can no longer participate in isometric force generation, but still hydrolyze ATP at the rate similar to that of native myosin heads. According to the dogma of Lymn-Taylor Scheme, however, myosin heads should hydrolyze ATP by repeating weak to strong actin binding transition. Result (4) obviously contradicts this dogma, since the strong myosin head binding to actin should produce internal resistance against muscle fiber shortening to result in decrease of Vmax.
Then, the question arises why myosin heads of antibody-attached myosin molecules do not produce internal resistance against muscle fiber shortening, while they repeat cyclic attachment-detachment with actin filaments, which has been believed to be coupled with the weak to strong actin-myosin binding transition? The puzzling question can only be accounted for by assuming that, when anti-S-2 antibody attaches to myosin molecules, their myosin heads no longer bind strongly to actin, but still hydrolyze ATP at the rate similar to that of myosin heads in native myosin molecules; myosin heads weakly binding with actin may readily be detached from actin with small vibrations applied to muscle fibers and also with myofilament sliding caused by a limited number of myosin heads in myosin molecules, which remain unbound with the antibody. The above explanation implies that, contrary to the generally believed dogma, ATP hydrolysis by myosin heads goes on without formation of strong actin-myosin binding. At present, we have no definite idea concerning the mechanism with which communication Figure 11: Relation between muscle fiber stiffness and Ca 2+ -activated isometric force in the presence of anti-S-2 antibody (1.5 mg/ml). The stiffness versus isometric force curves A, B, C and D were obtained before, at 30, 60, and 90 min after application of antibody, respectively (Insets) [13].  is transmitted from myosin S-2 region to myosin head-actin interface. Much more experimental work is necessary to reach full understanding about the regulation of actin-myosin binding by myosin S-2 region.

Effect of p-phenylene dimaleimide on contraction characteristics and MgATPase activity of skinned muscle fibers
In 1986, we studied the effect of p-PDM, which binds with reactive sulfydrils in myosin head distal region to produce loss of its ability to bind with actin filament and to hydrolyze ATP [14], on contraction characteristics and MgATPase activity of skinned muscle fibers [15] with the following results: (1) Three different species of myosin molecules are present as a result of partial p-PDM modification, i.e. two head native-, one head inactivated-, and two head inactivatedmyosin molecules, and their fractions to the total myosin molecules can be obtained by simple probabilistic calculation based on the total fraction of p-PDM-inactivated myosin heads, [14] (Figure 14). (2) Isometric force is only produced by two head native-myosin molecules, since the force is proportional to the square of (1-q) (Figure 15, left); (3) Mg ATPase activity is produced by both two head native-and one head native-(i.e. one head inactivated)-myosin molecules, since it is proportional to (1-q); and (4) One head native myosin molecules contribute not only to Mg ATPase activity, but also to internal resistance against muscle fiber shortening, as evidenced by the decrease of V max with decreasing isometric force (Figure 15, right); (5) Two head inactivated-myosin molecules do not contribute to both muscle fiber stiffness and internal resistance against fiber shortening, being consistent with the result of HMM solution experiments [14]. Result (4) indicates that, in one head native-myosin molecules can hydrolyze ATP at the normal rate, but cannot produce force (and shortening) by some steric hindrance caused by inactivated head in the same molecule. In this case, native myosin heads in one head inactivatedmyosin molecules hydrolyze ATP passing through the weak to strong actin-binding transition, which is no longer coupled with power stroke.

Further evidence for the regulation of actin-myosin binding strength by myosin S-2 obtained from in vitro motility assay experiments
To obtain further evidence for the regulation of actin-myosin binding strength by myosin S-2 region, we constructed an in vitro force-movement assay system by using a centrifuge microscope [16]. As illustrated in Figure 16, the centrifuge microscope consisted of a rotor and a stroboscopic device to observe and video-record ATPdependent movement of myosin-coated polystyrene beads along wellorganized actin filament arrays (actin cables) in a giant algal cell, which were mounted in a cuvette fixed on the rotor. Without application of centrifugal force, beads exhibit ATP-dependent movement along actin cables with a constant velocity irrespective of whether they are coated with untreated myosin (control beads) or coated with myosin previously treated with anti-S-2 antibody (1 mg/ml). This result seems to correspond to the result that Vmax of muscle fiber remains unchanged despite marked reduction of isometric force ( Figure 13). If centrifugal force is applied to the moving beads in the direction opposite to that of bead movement, the velocity of bead movement decreases with increasing centrifugal force, which serve as external load to the moving beads. The magnitude of centrifugal force required to stop and detach the moving beads are much larger for control beads (10.2 ± 4.8 pN, mean ± SD, n=8) than beads coated with antibody-treated myosin (1.1 ± 0.8 pN, n=7) [17]. This result constitutes additional evidence for the regulation of actin-myosin binding strength by myosin S-2 region.

Shortcomings of the swinging lever arm hypothesis
At the end of this review article, we will explain how the swinging lever arm hypothesis has been constructed, together with its shortcomings, As illustrated in Figure 17, a myosin head is like a tadpole in shape, consisting of oval shaped catalytic domain (CAD) and rod-like lever arm domain (LD), connected via small converter domain (COD). The LD is further connected to myosin filament backbone via myosin S-2 (not shown). Rayment et al. [18,19] prepared crystal of truncated myosin head (myosin S-1, obtained from slime mold Dictyostellium), in which most LD domain has been removed except for its base, and studied nucleotide-dependent structural changes in the myosin head crystal. Since it is not possible to actually bind nucleotides, ADP and ATP, to crystals, they used inorganic compounds as analogs of ADP and ATP. The nucleotide-dependent structural changes in the truncated myosin head were studied intensively by a number of investigators [20][21][22]. In addition to localized nucleotide-dependent structural changes around the nucleotide-binding cite in the CAD, they found that the base of the LD remaining in the truncated myosin exhibited rotation by ~60°. This rotation of the remaining LD base was taken to represent pre-and post-power stroke states of myosin head, Figure 14: Three different species of myosin molecules and their respective fractions expressed by the total fraction of p-PDM-inactivated myosin heads, q, calculated by assuming independent reaction of two myosin heads in the same myosin molecule with p-PDM. Open and filled heads represent native and p-PDM-inactivated heads, respectively [15].    assuming that the same rotation occurs if the whole LD is connected to truncated myosin. The above rather bold assumption indicates that the amplitude of rotation of at distal region of whole LD is enough to produce myosin power stroke in the direction parallel to myosin filament axis. Figure 18 is a diagram illustrating myosin head power and recovery strokes caused by active rotation of the CAD around the COD, as postulated by the swinging lever arm hypothesis.
Although the swinging lever arm hypothesis is now cited in every textbook as if it is an established fact, there are several shortcomings in the course of constructing this hypothesis. (1) To construct this theory, crystals of slime mold myosin was used instead of vertebrate skeletal muscle myosin because of the easiness to prepare crystals. Rather surprisingly, myosin head crystal structures were almost the same between slime mold and vertebrates despite the fact that slime mold differs so far from vertebrate animals. It seems possible that, due to close packing of myosin heads arising from crystal formation, functionally important differences in crystal structure between the two animals might be eliminated. (2) It is not well proved that the inorganic compounds used are actually good analogs of ADP and ATP. (3) Validity of the bold assumption to postulate rotation of the whole LD, based on the observed rotation of the remaining LD base in truncated myosin head is unclear. It seems rather unlikely that the small structures in the COD region can generate torque enough to produce myosin head power stroke in the myofilament-lattice, on which large external loads are imposed. The assumption that small structures in the COD region can actively rotate massive COD and CAD, with large loads applied to them, reminds me of an idiom, "The tail swings the dog". Such phenomena may take place in isolated molecules in solution, but are unlikely to occur in the 3D myofilament-lattice supporting large external loads. (4) It can be seen in Figure 18 that, during power and recovery strokes, change in angle takes place not only at the CAD-LD junction, but also at the LD-S-2 junction. We do not understand the reason why the angle change at the LD-S-2 junction can be regarded to be passive in nature. (5) In constructing the swinging lever arm hypothesis, crystallographists appeared to pay no attention to possible role of S-2 region in producing myosin head movement, and ignored extensive work made by Harrington et al. though they published papers in mainstream journals, which do not allow one to make an excuse to overlook them. We reserve further comment about the issues stated above.
Meanwhile, Muhlrad et al. [23] reported that chemical modification (trinitrophenylation) of reactive lysine residue in myosin head COD eliminated ATP-dependent in vito actin-myosin sliding and actinactivated MgATPase activity. They interpreted these results as being due to mechanical collision of structures in the COD to result in the loss of lever arm rotation around the COD. Contrary to their report, however, we found that a monoclonal antibody attaching around reactive lysine residue in the COD [24] had no effect on Ca 2+ -activated muscle contraction, while the antibody eliminated ATP-dependent in vitro actin-myosin sliding [25]. Generally speaking, binding of antibody to its epitope is reversible, while chemical modification is irreversible and may cause structural changes around the chemically modified region. Therefore, the report of Muhlrad et al. may not necessarily support the swinging lever arm mechanism, while our work may be taken to preclude it.

Conclusion
In this short review article, we presented evidence for the essential role of myosin S-2 region, which connect myosin heads to myosin filament backbone, especially its ability to regulate actin-myosin binding strength. At present, it is a mystery how myosin-S-2 transmit information to the actin-myosin interface at the distal region of myosin head CAD. In this article, we have also shown that the dogma in the Lymn-Taylor scheme is no longer valid: cyclic actin-myosin interaction coupled with ATP hydrolysis can take place in muscle without strong myosin head-actin binding when anti-S-2 antibody binds with S-2 region. In general, we hope muscle investigators, especially young ones, to realize that textbook view is convenient but not necessarily reliable, and start challenging a number of mysteries still remaining in the molecular mechanism of muscle contraction.