Muscle contraction is produced by the intermittent andasynchronous “working strokes” of many individual

myosin molecules that “row” the “thin” actin filaments pastthe “thick” myosin filaments. Actomyosin cross bridges areformed by the “heads” of the myosin II molecules that pro-trude from the shaft of the myosin thick filament and interactwith actin filaments in a cyclic manner, hydrolyzing a singleATP molecule in each cycle. Muscle myosin II molecules arenonprocessive molecular motors, i.e., ones that are releasedfrom actin after each catalytic cycle. In contrast, some non-muscle (or “unconventional”) myosins are processive motorsthat undergo multiple catalytic cycles and mechanical stepsfor each diffusional encounter with actin. The myosin super-family consists of at least 18 different classes distributed acrossplant and animal kingdoms and with great diversity of cellularfunctions (17). Reflecting these functional differences, there isconsiderable sequence and structural diversity in the tail partof the molecule, whereas the motor domain or head of themolecule is well conserved. It is assumed that throughout themyosin family the basic mechanism of movement and forceproduction is the same and occurs, as already mentioned, bythe cyclical interaction of the myosin motor domain with F-actin coupled to the breakdown of ATP.

Of all of the different members of the myosin family, mus-cle myosin II is surely the best studied. It provides a paradigmfor studying the structure/function relationships and how con-formational changes might generate movement and force.Here we will focus mainly on our current understanding ofthis myosin that has been optimized for a variety of contrac-tile functions, from the rapid repetitive contraction cycles ofinsect flight muscles to the extremely slow contractions oftonic smooth muscle. In addition, we will also touch on somesingle-molecule studies with unconventional myosin I. Therecent discovery of a two-step working stroke of myosin I givesimportant insight into the mechanism of cross-bridge move-ment and force production. These experiments (described inRef. 19) might offer insight into the molecular mechanism thatallows force maintenance to be economical from an energeticpoint of view, in particular in a state of tonic contraction suchas “latch” and “catch” of vertebrate and invertebrate smooth

muscles. Recently developed single-molecule mechanicaltechniques have made it possible to address the mechanism ofmovement and force production during a single cross-bridgecycle in an unequivocal way.

Structure of myosin II motors

Muscle myosin II heads [proteolytic subfragment 1 (S1)]consist of 1) an NH2-terminal catalytic (or motor) domain con-taining the actin-binding sites and the ATPase catalytic site and2) a neck region formed by an extended �-helix, to which arebound the essential and regulatory light chains (Refs. 7 and15; cf. Fig. 1). S1 retains all of the motor functions of myosinin vitro, i.e., the ability to produce motility and force. Further-more, limited proteolysis breaks S1 into three fragmentsnamed after their apparent molecular weights: 25 kDa (NH2-terminal), 50 kDa (middle), and 20 kDa (COOH-terminal).The crystal structure of S1 shows that these proteolytic frag-ments, rather than representing subdomains of S1, mark thepositions of two flexible loops (loop 1 and loop 2). It transpiresthat these loops are highly variable among myosin classes andare important in determining the biochemical and perhapsmechanical properties of different myosins.

A long �-helix, stabilized by the two noncovalently boundlight chains, projects from the ATPase catalytic site toward theCOOH-terminal tail of the molecule. It has been suggestedthat this “regulatory domain” (neck domain) acts as a “leverarm” structure that plays a central role in the production ofmechanical force and movement during the working stroke ofthe cross bridge. To date, several crystallographic structures ofchicken and scallop fast skeletal muscle myosin and verte-brate smooth muscle myosin with various nucleotide analogsbound in the nucleotide-binding pocket have been solved. Onthe basis of these crystal structures, it has been hypothesizedthat the cross-bridge working stroke is produced by an angu-lar movement of the myosin regulatory domain about a ful-crum in the so-called converter region of the myosin head. Theregulatory domain therefore acts as a lever to amplify smallstructural changes that take place close to the catalytic site inthe so-called switch I and II regions (Refs. 1 and 7; cf Fig. 2).

2130886-1714/02 5.00 © 2002 Int. Union Physiol. Sci./Am. Physiol. Soc. News Physiol Sci 17: 213�218, 2002;www.nips.org 10.1152/nips.01389.2002

Molecular Motors: Force and Movement Generated by Single Myosin II MoleculesCaspar Rüegg,1 Claudia Veigel,2 Justin E. Molloy,2 Stephan Schmitz,2 John C. Sparrow,2

and Rainer H. A. Fink1

1Department of Physiology and Pathophysiology, University of Heidelberg, INF 326, D-69126 Heidelberg, Germany; and 2Department of Biology,University of York, York YO10 5YW, UK

Muscle myosin II is an ATP-driven, actin-based molecular motor. Recent developments in opticaltweezers technology have made it possible to study movement and force production on the single-molecule level and to find out how different myosin isoforms may have adapted to their specificphysiological roles.

Subnanometer rearrangements at the active site are geared upto give a displacement of ~5�10 nm at the end of the leverarm. This lever movement would then drive a displacement of

the actin filament relative to the myosin head, and by deform-ing internal elastic structures it would also produce force (8).Part of that movement has been seen in cryoelectronmicroscopy studies of smooth muscle myosin bound to actin(20). An overall displacement of ~10 nm was estimated fromX-ray crystallographic studies of the motor domain of smoothmuscle subfragment 1 crystallized in the presence andabsence of ATP analogs (5). However, since crystal structuresof the actomyosin complex are still elusive, we do not knowthe effect of actin binding to myosin on its head structure ingeneral and, specifically, on the conformational change of thelever arm occurring during the working stroke of cycling crossbridges.

The cross-bridge cycle

During each cyclical interaction of myosin with actin, onemolecule of ATP is hydrolyzed by the myosin head into ADPand inorganic phosphate (Pi). Myosin binds to actin with theproducts of hydrolysis bound in the catalytic site to form anactomyosin-ADP-Pi complex (cross-bridge attachment). Thesubsequent, actin-activated release of Pi and then ADP is asso-ciated with a conformational change of the myosin head � theworking stroke � that generates movement. Binding of a newATP molecule to myosin causes the actomyosin rigor complexto dissociate (cross-bridge detachment), and subsequent ATPhydrolysis resets the original myosin conformation (“recoverystroke”) for another biochemical and mechanical cycle. Therowing action of cross bridges requires that the myosin affin-ity for actin is strong during the working stroke but weak dur-ing the recovery stroke.

Single-molecule mechanics

Development of single-molecule techniques over the past10 years have made it possible to study force and movementproduced by myosins in great detail. Most single-moleculemechanical studies of actomyosin use an optical tweezers-based transducer. The apparatus is usually built around a flu-orescence light microscope and uses near-infrared laser lightto produce the optical tweezers. Finer and colleagues (6) usedthis method to measure the piconewton forces and nanome-

News Physiol Sci • Vol. 17 • October 2002 • www.nips.org214

FIGURE 1. Ribbon representation of the myosin subfrag-ment 1 (after Ref. 15). The green, red/grey, and blue seg-ments represent the 25 kDa, 50 kDa, and 20 kDa segmentsof the heavy chain. The yellow and magenta parts corre-spond to the essential and regulatory light chains.

FIGURE 2. Hypothetical model of the swinging lever arm (after Refs. 1 and7). Top: myosin head attached to actin with the lever arm in pre-workingstroke orientation. Bottom: myosin lever arm in post-working stroke orienta-tion. Reprinted with permission from Ref. 7 [Annual Review of Biochemistry,Volume 68 © 1999, by Annual Reviews (www.AnnualReviews.org)].

ter displacements produced by single actomyosin interactions.In these experiments, small numbers of myosin molecules aredeposited on a solid support, e.g., a small (~2 �m diameter)silica bead, fixed to the surface of the microscope coverglass.A single myosin head is then brought into the vicinity of anactin filament, bound at either end to a plastic bead (~1 �mdiameter) that is held and positioned by laser traps (see Fig. 3).The position of these plastic beads is monitored by projectingtheir images onto a photodetector. The data trace in Fig. 3shows the movement of one of the plastic beads parallel to theactin filament axis during such an experiment. The bead-actin-bead dumbbell exhibits pronounced thermal motion that isoccasionally interrupted by intervals of reduced thermalmotion. It is now generally agreed that intervals of “highnoise” represent periods of Brownian motion when myosin isnot attached to the actin filament, whereas “low-noise” inter-vals identify periods of cross-bridge attachment to actin (13).In the presence of ATP, single myosin heads attach to actin,perform a working stroke, and then detach after a stochasti-cally distributed time interval that is dependent on the ATPconcentration in solution. Some aspects of these experimentsare analogous to single-channel recordings using patch clampmethods, in which the kinetics and conductivity of a channeltype can be determined from ensemble averaging of channelopening events. Both the kinetics and the size of the workingstroke of a cross bridge can be determined from the single-molecule mechanical data. Using ensemble averaging and so-called forced oscillations to improve time resolution to ~1 ms,the working stroke of rabbit fast skeletal muscle myosin II wasmeasured to be ~5 nm and was completed within 5 ms (Fig.4, bottom).

Using rather different approaches for measurement anddata analysis, most groups have confirmed a ~5 nm workingstroke for mammalian fast skeletal myosin S1 (e.g., Refs. 6, 13,18, and 19), whereas a 10- to 12-nm working stroke was esti-mated from mechanical studies of single muscle fibers (cf. Ref.8). In contrast to these findings, Yanagida’s group has claimedthat attached cross bridges may ratchet along actin sites in aseries of up to seven steps, slipping along the actin filamentsover 5�30 nm for each cross-bridge encounter (12), thus ques-tioning the lever arm hypothesis altogether.

The lever arm hypothesis of the cross-bridge workingstroke

If the lever arm hypothesis is correct, then the workingstroke size should depend linearly on lever length. Recentstudies in which the length of “engineered” lever arms hasbeen altered (16) are consistent with the idea that the regula-tory domain (neck domain) functions as a lever arm that tiltsduring the working stroke relative to the motor domain (andthe actin filament). But these experiments also suggest that thefunctional length of the lever arm may be somewhat longerthan the length of the neck domain. Thus comparing the stepsize produced by myosin constructs with shortened or elon-gated artificial neck domains and using the linear relationshipbetween displacement and lever arm length, extrapolation tozero step size placed the origin of the lever arm rotation >2

nm inside the motor domain (16).Furthermore, tomographical reconstructions of electron

micrographs of insect flight muscle that had been quicklyfrozen during contraction suggest that the working stroke iscomposed of a combination of tilting (rotation) of the motordomain on actin and lever arm swinging through ~35�. Also,a surprisingly large working stroke was found in single-mole-cule mechanical studies with proteolytically shortened rabbitskeletal S1 (14). All of these results are in agreement with theidea that, in addition to the lever arm motion, part of the work-ing stroke might be caused by a rotation of the motor domainabout its contact with actin, as originally proposed by H. E.Huxley, A. F. Huxley, and R. Simmons (see Ref. 8 for review).

One of the current challenges in single-molecule mechani-cal experiments is to resolve subtle details of the separatemechanical events that might occur during the early phases ofthe cross-bridge attachment, specifically during Pi and ADPrelease. So far, the time resolution has been insufficient to pro-vide insight into the phase between muscle myosin II attach-ment to actin and the end of the working stroke. So it remainsunclear whether the working stroke of the lever arm of fastskeletal muscle myosin is produced in a single step or in twoor more discrete substeps, as previously suggested (see Ref. 8for review). The experimental conditions are more favorable,however, if extremely slow myosins, specifically myosins with

News Physiol Sci • Vol. 17 • October 2002 • www.nips.org 215

FIGURE 3. Top: fluorescence image of 2 labeled plastic beads (1 �m diame-ter), each held in an optical tweezer. Between the beads, a fluorescentlylabeled actin filament is stretched out. Middle: arrangement of the single mol-ecule mechanical experiment. Bottom: time course of displacements of theactin filament in parallel to the actin filament axis. Displacements were meas-ured by observing the position of one of the plastic beads over time. Note theintervals of reduced Brownian motion, which indicate attachment of a crossbridge (see Ref. 13).

slow kinetics of ADP release, are used. Indeed, recent studieswith slow, unconventional myosin I, which is a single-headedmyosin, give insight into the details of this mechanism.

Myosin I produces its working stroke in two steps

Single-headed brush border myosin I from the microvilli ofintestinal epithelia and the 130 kDa myosin I from liver areslow myosins in terms of their ATPase kinetics and in vitromotility velocities (3). The slow kinetics and the long necklength of class I myosins with between three and six lightchain binding domains make them ideal for investigating thetiming of the working stroke. Using these myosins, it wasfound that the actin filament was displaced initially by ~6 nm(within a few milliseconds of myosin binding) and then, fol-lowing a time delay of 100�300 ms, a further movement of ~5nm occurred (Fig. 4, top and middle). The time intervalbetween the two substeps was independent of ATP concentra-tion, whereas the lifetime of the interaction following the sec-ond substep was shortened by increasing the ATP concentra-tion. This suggested that the phase following the second sub-step was terminated by ATP binding. Furthermore, the size ofthe second substep (~5 nm) was similar to the tilting of thelight chain domain of brush border myosin I induced by therelease of bound ADP, deduced from cryoelectron microscopyof myosin I-decorated actin filaments (10). It is tempting to

speculate that the two substeps of the brush-border myosinworking stroke are coupled to transitions between enzymaticstates, such as Pi release from the active site, followed by ADPrelease. If so, the intermediate attached state, lasting from theend of the first substep to the start of the second substep, prob-ably corresponds to a long-lived actomyosin-ADP state. Figu-ratively speaking, it probably represents a “caught” state of thelever arm, which, after performing the first phase of its work-ing stroke, may be blocked for a while before completing itsmovement and detaching from the actin filament.

In view of these results, the question arises as to whether atwo-step lever arm mechanism and the existence of long-lived,attached actomyosin-ADP states can also be demonstratedwith muscle myosin II. However, no tilting was observed incryoelectron microscopy reconstruction studies of skeletalmuscle myosin subfragment 1 in response to ADP (see reviewin Ref. 7), and mechanical substeps were not seen in single-molecule studies on skeletal muscle S1. One might argue thelack of such findings was merely due to the comparatively lowaffinity of fast skeletal myosin, so that movement associatedwith ADP release occurs too rapidly to be measured. Also,addition of ADP would not cause an ADP-bound state tobecome significantly populated and so movement wouldagain not be visible in cryoelectron microscopy studies. How-ever, studies by Yanagida’s group (in which chemical andmechanical states were observed simultaneously; Ref. 9)

News Physiol Sci • Vol. 17 • October 2002 • www.nips.org216

FIGURE 4. Displacement data from many crossbridge attachment events were aligned at thestarting and ending times, and the valuesobtained for each record at each time pointwere averaged. The ensemble average revealsthat the working stroke of two different class Imyosins [rat liver Myr-1a (top) and brush bordermyosin I (BBM1; middle)] is produced in twophases, whereas a two-step working strokecould not be resolved for skeletal musclemyosin (bottom). In class I myosins, the ATPdependence of the 2 phases of the workingstroke suggests that the first phase of the work-ing stroke is coupled to phosphate release,whereas the second phase is coupled to ADPrelease. The cartoons show how the regulatorydomains (containing either 2, 3, or 6 lightchains) might act as lever arms to amplify con-formational changes occurring in the catalyticdomain [Reprinted with permission fromNature (Ref. 19) copyright (1999) MacmillanMagazines Ltd.].

imply that there might be little or no connection between ADPrelease and lever arm movement (and force production) inskeletal muscle. Structural studies of slow (tonic) smooth mus-cle, however, indicate that the working stroke may well occurin two phases (20), as observed with myosin I.

Myosin motors in smooth muscle latch and catch states

Vertebrate smooth muscle myosin II probably does produceits working stroke in two phases (20), perhaps in a similar wayto brush border myosin I. In cryoelectron microscopy studies,two states separated by ~23� axial rotation of the lever armproduced by ADP release were observed (20). According toCremo and Geeves (4), opening of the ADP pocket to releaseADP might depend on mechanical strain of the cross bridges.With a high load on the strained cross bridge, ADP release isslower or even prevented and the lifetime of the attached crossbridges is therefore extended. Clearly, a prolonged attachedstate (in which cross bridges are occupied by ADP) would pro-vide a mechanism for economic tension maintenance in aholding function, e.g., in the so-called latch state of tonic ver-tebrate smooth muscle (11). This is important when we con-sider vascular smooth muscle that may maintain tension withlittle ATP consumption when withstanding a high blood pres-sure. In such a holding state, contractile force is supposedlysupported by slowly cycling “latchbridges,” dwelling for a pro-longed time in the intermediate attached state mentionedabove as bound ADP is released and replaced by ATP onlyslowly. However, ADP release can be accelerated by phos-phorylation of the regulatory myosin light chain, whichincreases the dissociation constant of the myosin-ADP com-plex and thereby shortens the intermediate attached state. Inthis way the latch state is terminated.

Recently Butler and colleagues (2) proposed that a long-lived intermediate attached state of myosin (i.e., an acto-myosin-ADP state) may also account for the so-called catchphenomenon observed in certain molluscan smooth musclesuch as the adductor muscles of bivalves, e.g., the scallop orthe oyster, that may keep the shells closed for many hours andeven days apparently without fatigue and barely any oxygenconsumption. Another catch muscle, the anterior byssusretractor muscle of the edible mussel Mytilus edulis, developsan enormous tension (up to 15 kg/cm2 cross section) whenstimulated electrically or with acetylcholine but relaxesextremely slowly or even fails to completely relax after cessa-tion of stimulation. Instead, one observes a remnant tensionknown as the catch, during which tension is maintaineddespite the fact that in this postcontractile state the cytosolicfree calcium concentration reaches resting values (<0.1 �M).In catch there is little if any energy turnover and the muscle isrigid, quasi in rigor, as if the cross bridges formed during thecontraction preceding the catch were “frozen” in the attachedstate. Indeed, there is evidence (reported in Ref. 2) suggestingthat, like the latch, the catch might well be due to the persis-tence of a locked actomyosin-ADP state or, for that matter, anintermediate attached state similar to that observed in myosinI. Thus actively contracting catch muscle may enter the catchstate if its cycling cross bridges are trapped at low free calcium

levels in a long-lived actomyosin-ADP state, stabilized per-haps by the high tension that pulls on the attached crossbridge. Furthermore, like the latch, the catch state too may beterminated by phosphorylation of a protein associated withthe myosin-containing myofilaments. However, the myosin-bound protein concerned is twitchin (a kind of mini-titin)rather than the myosin light chains that are involved in the reg-ulation of the latch state of vertebrate smooth muscle (2). Phys-iologically, this phosphorylation is brought about by proteinkinase A, for instance in an oyster when its adductor musclesare stimulated by the neurotransmitter serotonin liberatedfrom inhibitory nerves that eventually cause the shells to open.

Perspectives: functional and structural diversity ofmyosin motors

In the preceding sections, we have given a short overviewon how muscle myosin II isoforms are adapted to differentfunctions such as quick movement or holding. Some musclemyosins are specialized for rapid shortening and are charac-terized by fast cycling and very short (some milliseconds)attached states and possibly perform the lever arm movementin one step. Because fast cycling requires a high ATPase activ-ity, these motors are not economical for holding tension. Mus-cle myosins adapted for a holding function on the other handare necessarily slow. In this case the attached state, probablyan actomyosin-ADP state, is prolonged. Some of the slowunconventional myosins (e.g., brush border or liver myosin I)may even be capable of holding the cross bridge in a strain-dependent manner in an intermediate attached state for up toa second before completing the lever arm movement. Suchdiversities among myosins reflect functional specialization.The slow, single-headed liver cell myosin I, for instance, maybe involved in linking the cytoskeleton to the plasma mem-brane, whereas fast, double-headed skeletal muscle myosin IIinteracts with hundreds of actin partners to produce fast fila-ment sliding. At present, we know of at least 18 differentclasses of myosin in the plant and animal kingdom. Analysisof the human genome has shown that there are as many as 39genes representing 10 myosin families. In the future, thesemotor proteins may be expressed and purified, and theirmechanical properties could then be analyzed, specifically byusing novel assays combining different nanotechnology tech-niques on the single-molecule level. As we have seen, thesemethods permit the mechanical properties of differentmyosins to be measured in an unequivocal way (16, 19). Theimportance of such studies is underlined by the discovery thatmutations of some of these motor proteins can cause verysevere symptoms such as deafness or seizures as well as car-diac failure due to inherited cardiomyopathy.

References1. Block SM. Fifty ways to love your lever. Cell 87: 151�157, 1996.2. Butler TM, Narayan SR, Mooers SU, Hartshorne DJ, and Siegman MJ. The

myosin cross-bridge cycle and its control by twitchin phosphorylation incatch muscle. Biophys J 80: 415�426, 2001.

3. Coluccio LR and Geeves MA. Transient kinetic analysis of the 130-kDamyosin I (MYR-1 gene product) from rat liver � a myosin I designed for

News Physiol Sci • Vol. 17 • October 2002 • www.nips.org 217

maintenance of tension? J Biol Chem 274: 21575�21580, 1999.4. Cremo CR and Geeves MA. Interaction of actin and ADP with the head

domain of smooth muscle myosin: implications for strain dependent ADPrelease in smooth muscle. Biochemistry 37: 1969�1978, 1998.

5. Dominguez R, Freyzon Y, Trybus KM, and Cohen C. Crystal structure of avertebrate smooth muscle myosin motor domain and its complex with theessential light chain: visualization of the pre-powerstroke state. Cell 94:559�571, 1998.

6. Finer JT, Simmons RM, and Spudich JA. Single myosin molecule mechan-ics: piconewton forces and nanometre steps. Nature 368: 113�118, 1994.

7. Geeves MA and Holmes KC. Structural mechanism of muscle contrac-tion. Annu Rev Biochem 68: 687�728, 1999.

8. Huxley AF. Mechanics and models of the myosin motor. Philos Trans RSoc Lond B Biol Sci 355: 433�440, 2000.

9. Ishijima A, Kojima H, Funatsu T, Tokunaga M, Higuchi H, Tanaka H, andYanagida T. Simultaneous observation of individual ATPase and mechan-ical events by a single myosin molecule during interaction with actin. Cell92: 161�171, 1998.

10. Jontes JD and Milligan RA. Brush border myosin-I structure and ADPdependent conformational changes revealed by cryoelectron microscopyand image analysis. J Cell Biol 139: 683�693, 1997.

11. Khromov A, Somlyo AV, and Somlyo AP. Mg-ADP promotes a catch-likestate developed through force-calcium hysteresis in tonic smooth muscle.Biophys J 75: 1926�1934, 1998.

12. Kitamura K, Tokunaga M, Iwane AH, and Yanagida T. A single myosin

head moves along an actin filament with regular steps of 5.3 nanometres.Nature 397: 129�134, 1999.

13. Molloy JE, Burns JE, Kendrick-Jones J, Tregear RT, and White DCS. Move-ment and force produced by a single myosin head. Nature 378: 209�212,1995.

14. Molloy JE, Kendrick-Jones J, Veigel C, and Tregear RT. An unexpectedlylarge working stroke from chymotryptic fragments of myosin II. FEBS Lett480: 293�297, 2000.

15. Rayment I, Holden HM, Whittaker M, Yohn CB, Lorenz M, Holmes KC,and Milligan RA. Structure of the actin-myosin complex and its implica-tion for muscle contraction. Science 261: 58�65, 1993.

16. Ruff C, Furch M, Brenner B, Manstein DJ, and Meyhofer E. Single-mole-cule tracking of myosins with genetically engineered amplifier domains.Nat Struct Biol 8: 226�229, 2001.

17. Sellers JR. Myosins: a diverse superfamily. Biochim Biophys Acta 1496: 3�22, 2000.

18. Tyska MJ, Dupuis DE, Guilford WH, Patlack JB, Waller GS, Trybus KM, War-shaw DM, and Lowey S. Two heads of myosin are better than one for gen-erating force and motion. Proc Natl Acad Sci USA 96: 4402�4407, 1999.

19. Veigel C, Collucio LM, Jontes JD, Sparrow JC, Milligan RA, and Molloy JE.The motorprotein myosin-I produces its working stroke in two steps.Nature 398:530�533, 1999.

20. Whittaker M, Wilson-Kubalek EM, Smith JE, Faust L, Milligan RA, andSweeney HL. A 35-A movement of smooth muscle myosin on ADPrelease. Nature 378: 748�751, 1995.

News Physiol Sci • Vol. 17 • October 2002 • www.nips.org218

In Forthcoming Issue

Structure and Dynamics of the Fusion Pore in Live CellsBhanu P. Jena

Pulmonary Edema and Elevated Left Atrial Pressure: Four Hours and BeyondR. E. Drake and M. F. Doursout

Cerebrospinal Fluid Transport: A Lymphatic PerspectiveM. Johnson and C. Papaiconomou

Transport of Xenobiotics Across the Blood-Brain BarrierBruno Hagenbuch, Bo Gao, and P. J. Meier