Motion is fundamental to life. For centuries, the investigation of biological systems has involved the study of motion, from Antonie van Leeuwenhoek’s (98) early observations of “living animalcules, very prettily a-moving” to contemporary measurements of single kinesin molecules walking along microtubules (95). Steady improvements in biophysical instrumentation over the past decades have enabled the detection of biomolecular motions in vitro with near-atomic resolution, allowing the mechanisms governing motion in a wide range of biological systems to be investigated at the level of single molecules and individual catalytic turnovers. Some of these studies have investigated classical mechanoenzymes, also called motor proteins, such as kinesin, dynein, and myosin (3, 30, 57, 95, 112); rotary motors like F₁-ATPase and the bacterial flagellar motor (67, 90); processive nucleic acid enzymes like RNA polymerase (RNAP), DNAP, λ-exonuclease, and RecBCD (51, 70, 71, 73, 78, 105, 113, 114); and topoisomerases (34, 66, 92). Other measurements have probed vesicle trafficking (5) and membrane fusion (11). A large number of single-molecule studies have also probed conformational changes in macromolecules, such as the folding and strain response of RNA, DNA, proteins, and polysaccharides (25, 33, 39, 50, 53, 55, 58, 84, 87, 103, 116).

The broad application of single-molecule methods to study molecular motions reflects the ability of these techniques to address a variety of biologically important questions. The simplest of these include measurements of the fundamental step sizes of motors (1, 26, 37, 95, 108), issues related to motor processivity and load-bearing ability (3, 86, 94, 114), and the kinetics and thermodynamics of mechanical response and folding transitions (29, 55, 69, 104). However, deeper questions about the mechanism of molecular motion can also be addressed by single-molecule methods. These include the coupling between chemical and mechanical cycles (1, 20, 80), mechanisms of translocation (4, 26, 44, 110, 112), folding pathways and the properties of transition states (29, 55, 69, 104), and issues of enzyme stochasticity and population heterogeneity (27, 96, 106).

This review discusses the principles, capabilities, and practical implementation of five of the most widely used methods for making high-resolution measurements of the motion of single molecules: particle tracking, fluorescence resonance energy transfer (FRET), magnetic tweezers, force-mode atomic force microscopy (AFM), and optical traps. These methods can be divided into two broad classes. Centroid tracking and FRET fall into a class of techniques that passively track the motion of labeled molecules without applying significant external force. Magnetic tweezers, AFM, and optical traps fall into a second class of techniques that measure motion under the application of an external mechanical load.

All single-molecule methods rely on the attachment of labels to the molecule of interest, which either scatter light (such as micron-sized beads) or emit it (such as fluorophores or quantum dots), to allow visualization of molecular motions. For force-based methods, these labels must be attached to the molecule sufficiently well to withstand the forces being applied. Motion-reporter labels can be attached by using specific covalent interactions, specific noncovalent interactions, or nonspecific interactions. Covalent attachments based on amine-, thiol-, or carboxyl-reactive chemistries (41) can support large forces, on the order of 1 nN, but often involve harsh chemical treatments. Noncovalent attachments based, for example, on biotin-avidin (7), epitopes and antibodies, or metal chelation (79) are for this reason sometimes preferred. They can support forces from ~10 to ~300 pN (118). Nonspecific attachments such as pressure-induced (74) or charge-mediated (95) adsorption are simple to implement and can produce strong bonds, but the generally unknown orientation and mechanism of attachment can place limits on their application. For reasons of convenience or necessity, many single-molecule techniques have developed canonical linking chemistries, which may restrict their broader application to other specific systems of interest.

Aside from differences in the requirements of attachment chemistry, many of these techniques also operate in different regimes of spatial and temporal resolution, and they involve diverse levels of complexity in terms of the instrumentation required. Ideally, all of these characteristics should be matched to the requirements of a target biological system when choosing a particular experimental approach. A summary of features of the various methods is found Table 1 and discussed in greater detail below.

Table 1 A comparison of high-resolution, single-molecule measurement techniques

Figure 1 Centroid tracking of (a) particles and (b) single fluorescent molecules. Rotation can be monitored by measuring the motion of an attached, micron-sized particle. A single fluorophore attached to a protein of interest allows direct visualization of motion. (more ...)

Reporter particles Some of the earliest measurements of biomolecular motion were carried out using microscopic beads attached to the molecule (32, 83), and this approach remains popular and effective (Figure 1a). Reporter particles as diverse as micron-sized plastic or glass beads, gold particles, fluorescent beads, and fluorescent actin filaments or microtubules have all been used as suitable markers. These particles may be imaged with a microscope by using a variety of modalities, including brightfield, epifluorescence, differential interference contrast, back-focal-plane detection, or laser dark-field microscopy. The position of a molecule is inferred by tracking precisely the geometrical centroid (i.e., optical center of mass) of the particle image. The size, contrast range, and number of pixels in the imaging system determine the spatial resolution that can be achieved; typically, a digitized image allows the centroid to be located to within 1–2 nm (32), or about one tenth to one twentieth of a pixel, depending on the magnification. The temporal resolution is normally set by the image capture rate, but a fundamental limit is imposed by the characteristic relaxation time of the particle moving in its viscous medium.

Centroid tracking of particles is particularly well-suited to high-resolution measurements of rotary motion, as demonstrated by recent work on the dynamics of the F₁-ATPase and the bacterial flagellar motor. The F₁-ATPase is a part of the enzyme responsible for converting electrochemical potential into ATP; it can act as a rotary motor, producing torque through ATP hydrolysis. By attaching a fluorescently labeled, ~1-μm-long actin filament to the rotationally driven γ-subunit of F₁-ATPase, Kinosita and coworkers (67, 108) observed ATP-dependent rotation in discrete, 120° steps, confirming the Boyer model of ATP synthesis (12). Subsequent work increased the measurement bandwidth by replacing the long actin filaments with 40-nm gold particles measured by laser dark-field microscopy, reducing the viscous drag on the label. These improvements revealed substeps in the motion of F₁-ATPase: The 120° catalytic cycle consisted of a 90° step associated with ATP binding and a subsequent 30° step associated with hydrolysis product release (109). These particle-tracking investigations have provided impressive insights into the mechanisms of biomechanical coupling in the F₁-ATPase, as well as direct measures of its performance as a molecular motor.

The bacterial flagellar motor is another complex machine driven by ion gradients across a membrane. It is employed by bacteria to generate motility and can rotate flagella at speeds of hundreds of hertz (56). Berry and coworkers (77, 90) observed that the motor takes 26 equal steps per rotation by attaching 200-nm fluorescent or 500-nm nonfluorescent polystyrene beads to individual flagella and monitoring rotational motion at up to 40 Hz using epifluorescence or back-focal-plane detection, respectively. This work addressed a longstanding problem by implicating the FliG protein ring, a structural component of the flagellar motor with the same 26-unit periodicity, as the component responsible for torque generation.

These examples illustrate some of the practical considerations involved in particle tracking to measure motion. In particular, particle size has a pronounced effect on the measurement, because larger particles, although easier to visualize, limit the temporal resolution through hydrodynamic drag, which introduces damping. In addition to limiting the temporal resolution, drag on particles also applies forces to the molecules under study. This effect may be turned to advantage by varying the particle size to load the motor differentially, allowing force-velocity characteristics to be measured (67).

Fluorescent dyes When the perturbation induced by a reporter particle is too large, an effective alternative is to replace the particle with one or more fluorescent dyes attached to specific locations on the molecule and to visualize molecular motion by measuring the light emitted by the fluorophores (Figure 1b). By fitting the intensity profile of the fluorophores to the appropriate point spread function (e.g., an Airy function or Gaussian), the centroid can be determined with an uncertainty, δx, of δx = σ/√N, where N is the number of photons counted and σ≈λ/(2NA) is the diffraction-limited spot size (λ is the wavelength of the emitted light and NA is the numerical aperture of the measurement system). The resolution of centroid tracking is thus not limited by the Rayleigh criterion, but rather by counting statistics at short timescales and, importantly, by any residual baseline drift or motion of the system at longer timescales.

Centroid tracking of single, fluorescently labeled molecules has recently been applied to study the stepping mechanism of classical molecular motors. By attaching a fluorophore to one head of the homodimeric motor protein myosin V, which moves in 37-nm steps (76), Selvin and coworkers (110) observed directly that myosin V walks with a hand-overhand mechanism, in which each head takes alternating steps of 74 nm. Similarly, by attaching a fluorophore to a single head of a dimeric kinesin molecule, which moves with 8-nm steps (95), Yildiz et al. (112) observed alternating 16-nm steps, helping to show that kinesin, too, moves by an alternating hand-over-hand mechanism.

Successful single-molecule fluorescence tracking requires attention to a number of complicating factors. Fluorophores must be attached covalently in a region where they will not disturb the function of the molecule and also will not be quenched, necessitating significant structural knowledge of the target molecule. Reactive oxygen species can also rapidly bleach fluorophores. To mitigate photobleaching, oxygen-scavenging systems are often used to extend fluorophore lifetime (111). Finally, methods to block out-of-focus fluorescence and minimize the sticking of fluorophores to the cover glass surface can significantly enhance the signal-to-noise ratio. Under ideal circumstances, using total internal reflection fluorescence microscopy and oxygen scavengers, this type of centroid tracking can determine the position of a tagged molecular motor to within 1.5 nm every second over an interval of ~1 min (hence it has been dubbed fluorescence imaging with one nanometer accuracy, or FIONA) (111).

Centroid tracking of single fluorophores reaches its limits, however, when trying to probe internal molecular rearrangements or small motions that occur on the millisecond timescale. Such motions, especially small conformational changes, are more typically studied with a different fluorescence-based technique: FRET.

Figure 2 FRET mechanism and sensitivity. (Left) Excitation light (blue) excites the donor fluorophore ( yellow). When the acceptor fluorophore (red ) moves close to the excited donor, resonant energy transfer occurs, generating fluorescence centered at the acceptor (more ...)

Single-molecule FRET has been used to follow conformational changes in a wide variety of contexts, including the effects of calcium binding on calmodulin domains (13), the folding pathway of a cold-shock protein (81), the migration of Holliday junctions (59), the folding of telomeres (52), and transcription initiation (43). It has also been applied extensively to study the folding dynamics of single RNA molecules (115). RNA is a remarkably versatile macromolecule, capable not only of information storage, but also complicated three-dimensional folding and various catalytic functions. Chu and coworkers (116) probed the folding and catalytic activity of the Tetrahymena ribozyme, formed from a self-splicing group I intron, by observing FRET between dye pairs placed at the 3′ and 5′ ends of ribozymes tethered to a glass surface. These measurements supplied new insights into intermediate states and folding pathways. Similar measurements on the hairpin ribozyme, one of the simplest of catalytic RNAs, revealed complex interactions between folding and catalytic activity (117) and large heterogeneity in the folding behavior associated with loops formed in the ribozyme (96).

A number of practical issues require special attention to optimize FRET measurements. First, the donor and acceptor fluorophores must be selected carefully, so that both their emission and absorption spectra are reasonably well separated, yet with significant spectral overlap for efficient energy transfer. Commonly used fluorophore pairs with the appropriate spectral characteristics include Cy5 and Cy3, Cy3 and fluorescein, Alexa 488 and Alexa 594, blue-shifted GFP and red-shifted GFP, and even quantum dots (13, 38, 42, 81, 115). Other photophysics may strongly affect measurements: for example, blinking of the acceptor dye can mimic the desired FRET signals (38), and dye photobleaching may limit the duration of the measurement. Fluorescent molecules tend to generate a finite, characteristic number of photons on average before bleaching, typically less than one million (e.g., ~10⁵ for Cy5). In practice, the FRET efficiency can be measured to ~10% accuracy for every 100 photons, allowing ~10³ independent measurements with a maximum temporal resolution, set by fluorescence saturation, of ~1 ms (38).

Figure 3 The effect of external force on the energy landscape. The natural energy landscape is altered by applying a constant force to produce a perturbed energy landscape. This perturbation changes the height of the energy barrier, ΔE, by an amount equal (more ...)

The increased measurement resolution that accompanies the application of force follows from the load-dependent stiffness, k, of molecular systems. Most molecules have nonlinear elastic compliance, and stiffness tends to increase with tension. Through the equipartition theorem, any increased stiffness reduces the thermal fluctuations in position that limit resolution according to , where k_B is the Boltzmann constant, T is the absolute temperature, and dx_rms is rms displacement. In practice, the resolution increase is enhanced because the motion of the molecule is sampled only at frequencies below the characteristic viscous damping frequency for the molecular motion, f₀. The power spectrum, S, of thermally driven positional motion in a harmonic potential is a Lorentzian:

where f₀ = k/2πβ is the roll-off frequency and β is the drag coefficient of the trapped particle. The mean-squared thermal displacement for a measurement band Δf below f₀ is then given by dx² = (4βk_BTΔf)/k². Thermal fluctuations for a given measurement bandwidth can thus be reduced either by increasing the stiffness of the molecule (effectively reducing the amount of fluctuation) or by decreasing the drag coefficient β of the system (effectively moving fluctuations to higher frequencies where they become averaged by the measurement), as illustrated in Figure 4.

Figure 4 Brownian noise reduction in a limited bandwidth. Halving the viscous drag on the observed particle reduces mean-squared fluctuations in the measurement bandwidth by a factor of two. Stiffening the molecular handles by a factor of two reduces mean-squared (more ...)

To apply force in a controlled way, the load must first be calibrated. One approach uses Brownian motion of the force probe itself (e.g., AFM cantilever, trapped micro-sphere) to determine its stiffness, k_probe, either through the equipartition theorem or through the roll-off frequency of the associated thermal power spectrum (93). Alternatively, the response of the probe to a known force, such as viscous drag from hydrodynamic flow, can be measured. These approaches all have characteristic systematic errors that can affect calibration (8).

The application of force using compliant force probes can introduce spurious effects, owing to the elasticity of the molecule-probe system (Figure 5). Often, molecular handles (e.g., DNA or polypeptides) are attached to the molecule under study in order to apply force in the desired way. These handles, the molecule of interest, and the force probe itself all act as compliant springs. As a result, a given motion of the molecule, δx_mol, will produce a correspondingly smaller probe motion, according to δx_probe = δx_mol [k_mol/(k_mol + kprobe)], where the spring constants k_mol and k_probe refer to the molecular handle and probe, respectively. Moreover, motion of a spring-like probe necessarily changes the force applied to the molecule, complicating experimental interpretations. For these reasons, it is often convenient to employ a force clamp arrangement to maintain a constant load, either using active feedback (85) or passive methods (36). The use of a force clamp eliminates the need for compliance corrections due to changing system stiffnesses, and it can also improve the temporal resolution of displacement. However, force clamps using active feedback have a finite response time for engaging the feedback loop. This may complicate experimental interpretation if the response time, typically ~1–100 ms, is similar to the timescale for the motions being studied.

Figure 5 Measured distances are reduced by system compliance. (a) Force is applied by an elastic probe to a molecule (green) tethered by a molecular handle, and length changes are measured from the displacement of the force probe (blue cross). (b) The handle acts (more ...)

Figure 6 Magnetic tweezers. A superparamagnetic bead experiences a force in an inhomogeneous magnetic field and thereby applies load to a single molecule attached to its surface.

In addition to applying force, magnetic tweezers can also apply torque, as magnetic beads act as dipoles with a preferred orientation in the magnetic field. Comparatively high torques (up to ~1000 pN nm⁻¹) can be applied to molecules attached to a bead in a torsionally constrained manner. To detect rotational motion, the spherical symmetry of the magnetic beads must be broken, which is often accomplished by manipulating small clusters of beads such that their angular orientation is clearly discernable by imaging (35). Such methods have been used to study the mechanical response of DNA to force and torque (33, 88, 91), the motion of the DNA translocase FtsK (10, 78), and the molecular mechanism of topoisomerases (47, 92). Recent work from Ebright and coworkers (73) also used magnetic tweezers to monitor the formation of a transcription bubble by RNAP with single-base precision through the generation of plectonemes in the DNA template. This experiment, along with single-molecule FRET studies (43), proved that transcription initiation involves scrunching of the DNA template. Magnetic tweezers typically have fairly limited temporal resolution mainly because of the viscous drag associated with the relatively large magnetic beads used (~1–4 μm in diameter). Although unmatched for producing torque (proportional to B), magnetic tweezers can apply only limited force (proportional to B), and their axial and lateral position resolution is not as high as can be achieved with some of the other methods. However, they offer several advantages: Magnetic tweezers provide a stable platform for measuring slow molecular processes involving both force and torque; they avoid completely the use of potentially damaging photon fluxes; they can harness the formation of plectonemes in DNA to amplify the motion of nucleic acid–based enzymes (19); and they involve relatively simple, straightforward instrumentation.

Figure 7 Atomic force microscopy. A cantilever exerts tension on a molecule of interest attached to the tip. Motions are measured by recording with a position-sensing detector the deflection of a laser beam reflected off the cantilever, and force is modulated (more ...)

AFM has been applied with success to study unfolding in a number of proteins, including ankyrin (50), spectrin (75), barnase (9), and GFP (25), as well as structural rearrangements in polysaccharides (58), by measuring changes in the extension of the molecules as these denature under external load. The most extensive investigations have been carried out for the protein titin, an enormous, filamentous molecule from muscle sarcomeres that carries multiple-tandem-repeat domains of several distinct types and is believed to furnish passive elasticity to muscle tissues. Several groups characterized the nanomechanical properties of single titin molecules, including the unfolding force and key features of the folding landscape (74, 102). They also compared their results to denaturant studies of unfolding (17) and reconstructed titin’s role in muscle elasticity from single-molecule measurements of the individual structural components (53). The refolding of recombinant polyproteins, consisting of multiple repeats of a single protein domain of interest, has also been studied by Fernandez & Li (29), who found, for example, that polyubiquitin refolds in a complex energy landscape. Measurements on both polyubiquitin (15) and poly-GFP (24) also showed that the unfolding pathway depends on the location at which force is applied to the molecule.

Molecules with tandem repeats such as titin and polyprotein constructs are well suited for studies by AFM. Despite the random attachment points generated by the nonspecific binding to the probe, several repeats in series are generally placed under tension. The resulting regular array of characteristic sawtooth patterns observed in the force-extension curves, as each repeat unfolds, permits single molecules to be distinguished from multiple tethers or from various nonspecific interactions with the cover glass surface. Proteins that do not naturally have a tandem-repeat structure may be studied by forming recombinant chimeras between the proteins of interest and previously characterized, tandem-repeat handles (e.g., consisting of titin immunoglobulin-like domains or concatenated ubiquitin domains), which are then adsorbed nonspecifically to the tip (9). As a practical issue, care must be taken when designing single-molecule AFM studies to minimize potential complications arising from the nonspecific nature of the attachment to the AFM tip, the use of chimeric concatamers, and potentially complex interactions between the probe tip and the sample surface [via charge interactions, water exclusion, and other effects (118)].

Sources of noise affecting AFM measurements include motions of the sample with respect to the tip (e.g., acoustic coupling, thermal expansion, piezo creep) and optical interference effects between the tip and the surface. With careful isolation, instrument design, and alignment of the detection laser, such noise sources can be minimized, allowing intrinsic instrumental noise to be kept below the level of Brownian motion. However, because the tip of the cantilever is relatively large (hundreds of microns), viscous damping effects can limit detection bandwidth (see Figure 4). Because of the high cantilever stiffness, AFMs can be used to exert considerable force (between 10 and 10,000 pN), including forces sufficient to rupture covalent and high-affinity, noncovalent bonds. AFMs can therefore be used to measure the forces of ligand-substrate interactions for important macromolecular linkages, such as biotin-avidin, biotin-streptavidin, fluorescein/antifluorescein, and IgG/protein G (118), as well the unfolding forces of stable proteins (16) and the strength of disulfide bonds (101).

Figure 8 Comparison of the (a) surface-based assay, (b) dumbbell-based assay using one optical trap and a micropipette, and (c) dumbbell-based assay using two optical traps. Force is recorded by measuring the displacement of the bead from the center of the optical (more ...)

The position of the bead can be tracked by collecting the laser light scattered by the bead, either from the trapping laser itself or from a separate detection laser, using techniques such as optical trapping interferometry (23), back-focal-plane detection (99), or other variations (87). Trap stiffness is calibrated either from the variance or the frequency spectrum of the Brownian motion of trapped beads or from the response of a trapped bead to viscous drag forces induced by fluid flow (93). Stiffnesses are typically lower than those for AFM cantilevers, generally 0.01–1 pN nm⁻¹, and consequently the forces that can be applied are also lower, generally 0.1–100 pN.

Optical tweezers have been applied to study a wide range of biophysical problems. The characteristics and mechanisms of the motion of classical mechanoenzymes such as kinesin (4, 44, 80, 94, 95), myosin (3, 30, 62, 76), and dynein (37, 57, 97) have been investigated in depth. Optical traps have also been used to probe the motion and mechanisms of processive nucleic acid enzymes such as exonucleases and helicases (26, 70, 71), DNA translocases (72) and polymerases (105), and RNAP (1, 2, 21, 40, 63, 82). In beautiful studies of the packaging of viral DNA into the phi29 bacteriophage, Bustamante and coworkers demonstrated that the rotary portal motor of the bacteriophage can package DNA against high forces (86), and elucidated a minimal kinetic model of force generation (20). Looking at RNAP, Block and coworkers studied the frequent pauses during transcription elongation (63), finding that they were off-pathway and sequence dependent (40) but unrelated to secondary structure in the RNA transcript (21). High-resolution measurements of the motion revealed single base-pair steps spanning 3.4 Å and provided significant constraints on possible models of chemomechanical coupling (1).

Optical tweezers have also been used to investigate the elastic properties of nucleic acids (84, 87), to study nonequilibrium kinetics (54), to measure protein-DNA interactions (46), and to characterize the folding of nucleic acids (55, 69, 104) and proteins (18, 45). By measuring the force-induced folding and unfolding of a single nucleic acid hairpin with very high position resolution and stability, Woodside et al. (103) determined the shape of the folding energy landscape along the full length of the reaction coordinate and identified the effects of sequence changes.

When measuring motion with optical tweezers, the main instrumental consideration is the geometry to be used, which depends on both the system under study and the required resolution and stability. Surface-based assays, in which one end of the molecule is tethered to the trapped bead and the other end to the surface of a cover glass (Figure 8a), are the simplest. However, they are susceptible to relative drift of the surface and the trap (e.g., due to thermal heating of the objective, stage settling, laser pointing fluctuations, or acoustic noise coupling). By independently measuring the relative positions of the trap and surface, noise from surface drift can be reduced to ~1 Å (68). The pipette-tethered dumbbell assay (87), which involves attaching one end of the molecule to a bead held in the optical tweezers and the other to a bead held by a micropipette (Figure 8b), allows for versatile manipulation of the sample and increased position sensitivity. This geometry, however, is still subject to the same noise sources as the surface-based assay, because of the mechanical connection through the micropipette.

Using a second optical trap instead of the micropipette (Figure 8c) greatly reduces the mechanical noise, providing the most stable configuration (82). At this level, fluctuations in the relative positions of the laser beams induced by air currents contribute significantly to low-frequency noise. This noise can be suppressed by propagating the lasers through a gas with a low index of refraction, such as He (n = 1.000035) (1), or by careful isolation of the beam paths (61). The dual-trap dumbbell geometry also permits the simple implementation of an all-optical, passive force clamp to maintain constant force during the course of molecular motion, avoiding the complexities arising from finite-loop response times (36). Geometries with more than two traps are also possible (22).

With various improvements, optical traps have recently attained a resolution of 1 Å or better (at ~50 Hz) for the displacement of a trapped, micron-scale bead, which corresponds to determining position within the diameter of a single hydrogen atom. Dual-trap, independent-detector optical tweezers are complex instruments to produce, requiring at least three independent, diffraction-limited laser beams, some mechanism to control precisely the position of at least one trap at high bandwidth, and a low-vibration, temperature-stabilized environment for the apparatus. These devices can exert a large range of forces on small beads and have superior noise characteristics, but the required experimental geometry may be challenging to achieve in practice. Despite such limitations, optical traps are currently capable of measuring the motions of single biomolecules in real time over comparatively long timescales (minutes to hours) with atomic-level precision, which rivals the resolution attained by crystallographic methods.

Novel approaches to precision measurements of single-molecule motion continue to be developed. These new techniques include optical torque wrenches, which extend the functionality of optical tweezers to apply and measure controlled torque as well as force; various molecular rulers, which provide the means to measure motions and conformational changes on length scales both larger and smaller than those accessible by FRET; and combinations of force- and fluorescence-based techniques, which allow motions under load to be correlated with conformational changes.

An optical torque wrench (48) operates by trapping particles that are optically birefringent, i.e., possess either an anisotropic refractive index or a pronounced shape anisotropy. Such particles have a strong trapping axis, corresponding either to the slow optical axis or the long axis of the particle, which tends to align with the polarization of a linearly polarized trap beam. Rotation of the trap polarization exerts a torque on the particle as it attempts to maintain alignment with the polarization. The torque can be kept constant with a torque clamp, analogous to a force clamp, and the applied torque can be measured by detecting the change in angular momentum of the light scattered by the trapped particle. Torque wrenches should prove useful in studying systems with both rotary and linear components to their motion.

A number of spectroscopic molecular rulers are being developed to complement the capabilities of FRET. Single-molecule electron transfer (107), in which distance-dependent electron transfer to a fluorophore modifies the fluorescence lifetime, is capable of measuring distance changes in the Ångström and sub-Ångström regimes. Distances at the 1- to 3-nm length scale have been measured using single-molecule optical switching (6), in which the recovery kinetics of a single Cy5 fluorophore switched into a dark state by an applied laser beam depends upon the spatial proximity of a Cy3 fluorophore. At the other end of the distance scale, motions at a distance of up to 70 nm can be monitored with 1 nm resolution, by measuring shifts in the plasmon resonance frequency of metallic nanoparticle probes (89). This method also avoids the problem of photobleaching, but at the cost of using large probes.

Combinations of different single-molecule techniques promise to provide powerful tools that bring to bear simultaneously the advantages of individual techniques. One example is the combination of fluorescence and optical trapping (49), which would be useful to probe directly questions regarding the coupling of mechanical and chemical steps in enzymatic reactions. Such combinations may be improved by design changes that compromise the performance of the constituent techniques but ensure the success of the whole, such as the use of chopped trapping and excitation beams in a combined fluorescence/trapping apparatus to improve signal-to-noise ratios and increase fluorescence lifetimes (14).

The high-resolution methods of single-molecule investigation described here provide a rich toolbox of precision probes suitable for the study of a wide range of interesting biomolecular motions. Given a particular system of interest, selecting the most appropriate single-molecule technique involves several considerations. Among these are the temporal and spatial resolution needed to resolve any relevant molecular motions, the desirability of applying either force or torque to the molecule, the variety of alternative methods available for the attachment of chemical labels or mechanical handles in a way that preserves function, and an assessment of the complexity and cost of any apparatus required, which must often be constructed, tested, and calibrated. The many successful applications of the various tools for studying single-molecule motion inspire hope that other, less well-understood instances of motion, that most fundamental characteristic of life, are ripe for future investigation.