No hype in hyperspace

Main text

Targeted transport of intracellular vesicular cargos is mediated by myosin, dynein, and kinesin molecular motors operating on actin and microtubule cytoskeletal filaments (1, 2, 3). In this issue of Biophysical Journal, a new and powerful method is introduced to dynamically resolve the three-dimensional position and orientation of vesicles to help quantify and categorize these motions (4). The dynamic organization of the cytoskeleton is complex, with actin filaments and microtubules interacting throughout the cytoplasm. Vesicle movement, driven by motors, is characterized by travel at relatively constant velocity, and then stalling or diffusing randomly when the vesicle detaches from a filament, encounters an obstacle, or reaches a filament intersection (5,6). The quantitative characteristics of these motions can help to reveal their basis in regard to which motors are operating, how many are engaged, how they are regulated, how they interact with each other, and how their destinations are set. There is a large body of literature on tracking the positions of intracellular organelles at high temporal and spatial resolution to gain this understanding (reviewed in (7)). Although angular changes may signal numbers of motors linking the cargo to the cytoskeleton, the directions of their force vectors, and switching between motor types, rotation of vesicles has garnered less attention. This may be because of more complex microscopy and analysis or less robust signals that report the orientation.

A similar situation exists for tracking protein motions in the cell and in vitro. The positions adopted by labeled macromolecules inside of cells have been extensively investigated using high- and superresolution microscopy (reviewed in (8,9)). Distance measurements, for instance by fluorescence energy transfer, are also widespread (10). Less common is orientation measurement to determine relative domain motions or tilting of subunits relative to their substrate, some of which are crucial for their function. Spatial orientations of fluorescent probes and oblong light-scattering particles can be determined by polarized fluorescence, engineered point spread functions, pattern matching in defocused images, and differential interference microscopy (reviewed in (11).

Cheng et al. (4) monitored transferrin conjugated gold nanorods (GNRs), which are taken up by endocytosis into several cell types. GNRs seem to dependably track the position and angle of the vesicles, perhaps by being tightly wrapped in the endosomal membrane (12). Three position coordinates and two angular values amount to five-dimensional imaging (six-dimensional if time is included). Compared with absorption and emission from fluorescent probes, light scattering from GNRs has the advantages that are as follows: GNRs do not photobleach and they give high contrast. On the other hand, they are large, e.g., 30 × 60 nm, compared with individual protein domains which might restrict or slow rotations in that application.

To collect all of the signals, Cheng et al. built a darkfield microscope that separated the light scattered by the GNR into four separate images: two of them in focus and two purposely maintained at 0.9 μm out-of-focus by an extra lens in their pathway to the camera. The shapes of the defocused image spots convey orientation information because of the anisotropic scattering intensity that depends on the orientation of the GNR. They were kept at their in- and out-of-focus viewpoints by a feedback system driving a piezo-electric objective z axis scanner. Each of these spots was further split into two by a wedge prism near the objective back focal plane that deflected half of the objective’s collected light by a small angle to form a pair. Because of the effective sideways views of the GNR by the two halves of the light collecting cone, the y-positions at the camera varies with the z-position of the probe. Use of this parallax effect is similar to that used by Friedrich Bessel, who first estimated the distance to nearby stars in 1838 from their positions relative to more distal stars observed half a year apart when Earth had orbited to the opposite side of the sun (Fig. 1 A; (13)). A limitation of using parallax as the error signal for autofocus in the darkfield microscope is that only one GNR can be tracked and locked into focus during a given recording.

Figure 1.

Intuitive explanations of the depth and angular measurements in Cheng et al. (4). (A) Astronomical parallax. A nearby star (a) appears in a slightly different position relative to distal stars when viewed from the earth displaced by an orbital diameter. The difference in y-position of the target object between the two views (compare star a, solid light rays, vs. star b, dashed rays) is related to its distance (the horizontal axis in this drawing). In Cheng et al.’s setup, the two views are provided simultaneously by two halves of the microscope objective. They did not need to wait 6 months in between measurements. (B and C) Theoretical point spread function intensities calculated by Chen et al. (4), for scattering nanorods at the angles shown. The left columns in (B) and (C) are in-focus images; the right columns are defocused by 0.9 μm. Full-plane images are from the whole objective; half-plane images are from the two halves of the objective light collecting cone. The shapes of these images carry angular information extracted by comparing measured data with these patterns. (modified from (4) with permission). To see this figure in color, go online.

The authors present an impressive first-principle calculation of the four expected image shapes as functions of the defocus and the angle (Fig. 1, B and C) which are compared with the four experimental images to recover the GNR orientation. Splitting the scattered light into four spots does not overly degrade the signals because of the high intensity. The combination of biplane and parallax detection provides high speed (50 frames per second) and high precision (σ = 5–15 nm) of the x-, y-, and z-locations, which come from the in-focus image pair and high-resolution (∼2°) angular measurements of the polar (θ, relative to the z optical axis, similar to latitude) and azimuthal angles (ϕ, around the optical axis, the longitude) from the defocused pair. The angles were expressed unambiguously in a hemisphere defined by θ = 0 ≤ θ ≤ 90°, and ϕ = 0 ≤ ϕ ≤ 360°, which is the maximal angular range possible with a twofold symmetrical object like a GNR.

The limitation of angles determined within a hemisphere is intrinsic to an individual optical measurement unless other information is available. Even if the two ends of the probe are distinguished, any given probe measurement at (θ, ϕ) gives exactly the same image and polarizations as a probe at (180°–θ, ϕ ± 180°). When the probe is near θ = 90°, noise or real motions may carry it across the 90° equator. It might move away again into either hemisphere introducing ambiguity. This issue was circumvented in studies of myosin orientation in muscle fiber studies by applying quick stretches to the fibers, which cause sliding between the actin and myosin filaments and a predictable direction of tilting (14). The resulting angle changes distinguish which hemisphere contains the probe, either toward or away from the center of the sarcomere. Another way to resolve the twofold symmetry problem in some cases is to realize that the upper and lower hemispheres in the microscope are not the only choices for the range of detectable angles. If the probe changes angle less than 180° during the measurement, then a hemisphere, possibly tilted relative to the microscope axis, will contain the probe and there is no ambiguity. To find the unknown pole of this hemisphere for a set of measurements might depend on limited total motion between subsequent observations. The measurements may have a “director,” which is the predominant vector about which the probe moves (15). The angles of a subunit within a molecular motor walking reasonably straight along a cytoskeletal filament tend to be contained within a hemisphere defined by a director (16). Whether this feature applies to cellular vesicles, especially during free diffusion, is not clear, but Cheng et al. stuck to the upper hemisphere for simplicity.

Motions of endocytic vesicles were transiently categorized by Cheng et al., into active transport, restricted motions, and free diffusion according to the commonly used exponent α in the mean squared displacement equation, <r²> = A t ^α, where r is the distance the particle traveled at time t from its starting position. For free diffusion, the exponent, α, is equal to 1 and A = 6 D, where D is the diffusion coefficient. α is greater than 1 (e.g., ∼2) for active motion and less than 1 for restricted diffusion. Dynamics of the rotational motions further categorized periods of tethered rotation or restricted angle, presumably depending on number and cytoskeletal engagement of the motors on the vesicle surface. Examples of switching between these modes are shown in the paper. A fuller investigation of endocytic vesicle motion and other interesting objects should emerge in future work.

A puzzling result reported by Cheng et al. is that the fluctuations in the polar angle are apparently much greater than the azimuthal motions. An elongated object tends to rotate about its own axis more rapidly than perpendicular to that axis but in a homogeneous environment, free diffusion of the major axis is isotropic. The discrepancy between polar and azimuthal fluctuations might be a limitation of the method, for instance the incomplete range of measured polar angles, or it might be caused by anisotropy of the cytoplasm in the cells cultured in a single layer on a flat surface. It will be interesting to see if this observation holds up in three-dimensional culture.

Kaplan et al. (17) also measured two-dimensional position and angle of GNRs, using wheat germ agglutinin conjugated particles taken up and transported into axons of neuronal cells. The orientation was detected by two- or three-channel polarized light scattering and they also showed that the GNRs faithfully reported the endosome angles. They found that the vesicles aligned with the underlying microtubule axis, rotated markedly during reversals of transport direction, but not much during pauses. These observations seem to contrast with the properties of the GNRs used by Cheng et al. in a more compact cancer cell line. The complement of motors on endosomes and control of their engagement with cytoskeletal filaments are probably tuned to the specific requirements for transport in different cell types or cell compartments.

The clever optical and analytical methods reported in this paper should be straightforward enough to be duplicated for other cellular objects and by other labs interested in high-resolution positional and angular tracking. Future studies of this kind promise to help reveal the puzzling relationships between molecular motor activity and targeted transport of intracellular cargos.

Editor: Ahmet Yildiz.