(A) Illustration of the dynein dimer with associated chains and N-terminal GFPs.
(B) Schematic representation of the optical trapping assay (not to scale).
(C) Displacement of a single dynein molecule at 1 mM ATP in a fixed (nonfeedback) optical trap showing successive motor detachments and stalling events (trap stiffness: k = 0.055 pN/nm).The inset on the right side shows the stall force distribution (6.9 pN ± 1 pN; mean ± SD; n = 108).
(D) Stall force as a function of ATP concentration. Values are displayed as mean ± SD (n = 20–108).
(E) Velocity-force relationship at 1 mM ATP. All velocities were measured with the optical trap maintaining a constant load using feedback control, except the zero load velocity, which was measured by tracking the GFP-tagged motors in a single-molecule fluorescence assay. Values are displayed as mean ± SEM (n = 35–151).

(A) Stepwise backward movement of Dyn1_471kDa under 10 pN superstall force in the absence of ATP (trap stiffness: k = 0.055 pN/nm) (upper inset). The lower inset shows an example record of a microtubule-bound Dyn1_471kDa motor under stall and superstall loads in the absence of ATP. The motor is tightly bound in rigor for ~9 s under 7 pN stall force without detectable advancing steps and starts to step backward after the application of a 10 pN superstall force (trap stiffness: k = 0.063 pN/nm). The raw data are shown in black and the steps detected by the step-finding program in red.
(B) Optical trapping record of microtubule minus-end-directed movement of Dyn1_471kDa under –3 pN forward load in the absence of ATP (trap stiffness: k = 0.055 pN/nm). The trace segments (a, b and c) correspond to the trace sections indicated by the rectangular boxes.
(C) Forced-backward movement of Dyn1_471kDa under a simultaneous lateral load of 9 pN and a longitudinal backward load of 10 pN (trap stiffness: k = 0.07 pN/nm). The record shows the displacements of the trapped bead along the microtubule axis (x) and in perpendicular direction (y). The inserted trace segments (a and b) correspond to the trace sections indicated by the rectangular boxes.
(D) Histograms of step sizes for microtubule-minus-end-directed movement under –3 pN assisting load (left, n = 292) and microtubule-plus-end-directed movement under 10 pN rear-ward load (right, n = 332) in the absence of nucleotide.

(A) Example record of nonadvancing stepping under 3 pN rearward load.
(B) Rate constants k_f and k_b of nonadvancing forward and backward stepping as a function of load and ATP concentration. The rate constants (mean values) were obtained by cumulative-distribution analysis of the underlying dwell time data (Figure S12). The load dependence of the rate constant k_b at 1 mM ATP (black squares) can be expressed by an exponential function of the form exp(Fd/k_BT), with and d = 0.79 ± 0.05 nm. Error estimates were calculated as the SD of fit parameters derived from 200 bootstrap samples drawn from the underlying data set.

Mechanical pathways for force-induced backward (A) and forward (B) stepping and an ATP-dependent pathway for normal forward stepping (C). The key feature proposed for these pathways is a tension-sensing mechanism by the MTBD. In these models, forward deflection of the stalk (induced by external forward load [pathway B] or intramolecular strain provided by a power stroke [pathway C]) weakens the binding affinity of the MTBD in the rear head (indicated by the orange-colored stalks). This mechanism favors rear head detachment and thus helps to keep the dynein heads out-of-phase during continuous movement toward the microtubule minus end. Backward load potentially increases the microtubule-binding affinity of the MTBD in the front head (caused by a load-induced backward deflection of the stalk) (indicated by the red-colored stalk), which explains the large external loads required to induce backward stepping. The size of the α/β tubulin dimers and the length of the stalk and the diameter of the dynein ring are drawn to scale. See the Discussion for more details.

(A) Optical trapping record of forced-backward movement of Dyn1_471kDa under 10 pN super-stall force (force-feedback mode, gray-shaded area) following a motor run under increasing rearward load (nonfeedback, upper corner left, the red coordinate system indicates the corresponding force) in the presence of 1 mM ATP (trap stiffness: k = 0.056 pN/nm).
(B) Forward movement of Dyn1_471kDa under a –3 pN assisting load and 1 mM ATP (trap stiffness: k = 0.033 pN/nm).
(C) Velocity (absolute values) of Dyn1_471kDa movement at 10 pN rearward load (n = 54–77) and –3 pN forward load (n = 52–60), respectively, as a function of ATP concentration. Values are displayed as mean ± SEM.

(A–D) Histograms of steps sizes and example trace segments measured under 1 pN (n = 716), 3 pN (n = 1104), 6 pN (n = 1314), and 10 pN (n = 642) constant rearward load (force-feedback). The gray-shaded histograms correspond to the combined step size data and the red and blue histogram bars indicate the steps assigned to the advancing and nonadvancing modes, respectively. The raw stepping data are shown in black and the steps detected by the step-finding program in red (advancing mode) and blue (nonadvancing mode). The probability for taking a backward step in the advancing mode (, with being the probability for taking a forward step in the advancing mode) at 1, 3, 6, and 10 pN load is 0.26, 0.34, 0.4, and 0.75, respectively (calculated from advancing step size histograms shown in red). The probability p^non-adv for taking any step in the non-advancing mode (p^adv + p^non-adv = 1, with p^adv being the probability for taking any step in the advancing mode) is 0.19, 0.42, 0.5, and 0.31 at 1, 3, 6, and 10 pN, respectively.

(A) Compact and extended conformations of the dynein dimer may explain a wide variation in step size. In its compact state, the two dynein rings are restrained and located in close proximity and perhaps overlapping due to direct head-to-head “interactions” or a “zipping” of the proximal tail. The loss of physical interactions or an “unzipping” of the proximal tail might cause a less restrained extended conformation with an increased head-to-head distance.
(B) Diagram of constructs showing the dynein heavy chain truncations and tags.
(C) Force production and stepping behavior of GST-Dyn1_331kDa. Left: Schematic of the GST-Dyn1_331kDa motor. Center: Histograms of the combined step size data (gray-shaded), classified according to the advancing (red histogram bars) and nonadvancing modes (blue histogram bars) (3 pN force-feedback data, n = 670). Right: Stall force distribution of GST-Dyn1_331kDa (4.8 pN ± 1.0 pN; mean ± SD; n = 195).
(D) Force production and stepping behavior of GST-Dyn1_314kDa. Left: Schematic of the GST-Dyn1_314kDa motor. Center: Histograms of the combined, advancing and non-advancing step size data (3 pN force-feedback data, n = 518). Right: Stall force distribution of GST-Dyn1_314kDa (4.0 pN ± 1.1 pN; mean ± SD; n = 91).
(E) Force production and stepping behavior of the dynein construct GST-α2-Dyn1_314kDa with artificial linker elements. Left: Schematic of the GST-α2-Dyn1_314kDa motor. Center: Histograms of the combined, advancing, and nonadvancing step size data (3 pN force-feedback data, n = 457). Right: Stall force distribution of GST-α2-Dyn1_314kDa (5.2 pN ± 1.1 pN; mean ± SD; n = 80).