Retraction in Amoeboid Cell Motility Powered by Cytoskeletal Dynamics

See allHide authors and affiliations

Science  21 Nov 2003:
Vol. 302, Issue 5649, pp. 1405-1407
DOI: 10.1126/science.1089129


Cells crawl by coupling protrusion of their leading edge with retraction of their cell body. Protrusion is generated by the polymerization and bundling of filaments, but the mechanism of retraction is less clear. We have reconstituted retraction in vitro by adding Yersinia tyrosine phosphatase to the major sperm protein–based motility apparatus assembled from Ascaris sperm extracts. Retraction in vitro parallels that observed in vivo and is generated primarily by disassembly and rearrangement of the cytoskeleton. Therefore, cytoskeletal dynamics alone, unassisted by conventional motors, are able to generate both of these central components of amoeboid locomotion.

The crawling movement of metazoan cells responsible for tissue formation, wound healing, immune surveillance, and tumor metastasis is typically powered by their actin cytoskeleton. Locomotion requires both protrusion of the leading edge of the cell's lamellipod and retraction to pull its trailing cell body forward, coupled with modulation of adhesion to the substrate (1, 2). Analysis of several types of crawling cells, as well as analogous motile systems such as the actin-based rocketing motility of the bacterium Listeria monocytogenes, has indicated that protrusion is generated primarily by the localized assembly of branched networks of actin filaments along the leading edge of the lamellipod (26). Establishing the precise mechanism of retraction has been difficult, partly because there has not been an in vitro model system available in which to analyze this process, although it has been proposed that in some cells, retraction may involve actomyosin contraction (1, 710).

We report here reconstitution of retraction in vitro using the motility apparatus of sperm from the nematode Ascaris suum. This simple motile system, in which the components of the locomotory machinery are reduced to a minimum, has provided valuable insight into the mechanism of cell crawling (11). Locomotion in Ascaris sperm is indistinguishable from that of many other crawling eukaryotic cells, even though sperm use a cytoskeleton based on major sperm protein (MSP) rather than actin (12). In Ascaris sperm the MSP filament system is dedicated to locomotion and is organized specifically to generate cell movement. The MSP filaments are arranged into meshworks, called fiber complexes, that can be seen by light microscopy. Fiber complexes assemble at the leading edge, where protrusion occurs, then move rearward and are ultimately taken apart at the rear of the lamellipod, where the organelle-packed cell body is pulled forward. Although protrusion and retraction occur at opposite ends of the lamellipod, in nematode sperm their rates are balanced, so that the leading edge of the lamellipod and the cell body advance at the same speed (12). Protrusion and retraction can be uncoupled by modulation of intracellular pH. Under these conditions, protrusion stops but shortening of the fiber complexes continues as they are taken apart at the base of the lamellipod, creating tension in the cytoskeleton that pulls the cell body forward (13).

The MSP-based protrusion of the leading edge can be reconstituted in cell-free extracts of sperm (S100), and this in vitro system has shown that, as in actin-based cells, protrusion is powered by membrane-associated cytoskeletal assembly. Vesicles derived from the leading-edge membrane trigger the assembly of a fiber composed of a columnar meshwork of MSP filaments (14). Localized polymerization of filaments and their arrangement into a meshwork at the vesicle surface allow fibers to push their vesicle forward as they elongate in the same way that a column of cross-linked actin filaments pushes objects such as Listeria (3, 4, 6).

The localized cytoskeletal disassembly associated with cell body retraction in vivo does not typically occur in fibers assembled in vitro. However, we found that addition of Yersinia enterocolytica tyrosine phosphatase (YOP) to fibers growing in S100 generated retraction in the MSP motility apparatus (15). Thus, when fibers were perfused with S100 supplemented with YOP (1.3 U/ml) but without added adenosine 5′-triphosphate (ATP), they stopped growing and started to shorten (Fig. 1A, movie S1). This fiber retraction frequently continued until the fiber was less than 10% of its original length and at times until it was no longer detectable. Fibers shortened at 1.7 ± 0.7 μm/min, and this rate was not strongly dependent on fiber length (Fig. 1C). Shortening occurred along the entire length of the fiber and was accompanied by a progressive decrease in its optical density (OD). Previous work had shown that the loss of fiber OD is due to depolymerization of its constituent filaments (16). Fibers perfused with assembly buffer alone also lost OD but did not shorten extensively (Fig. 1, B to D).

Fig. 1.

Reconstitution of retraction in vitro. (A) Time-lapse sequence of phase contrast micrographs of a fiber assembled in S100 and perfused with S100 containing YOP (1.3 U/ml). The vesicle-bearing end of the fiber is at the top. The distance from each end of the fiber (white arrowheads) to a piece of debris attached to its side (black arrowheads) decreased with time, indicating that shortening occurred throughout the fiber. The OD of the fiber decreased as it shortened. In this example, the fiber lost 74% of it length and 71% of its OD in 19 min. (B) A similar sequence showing the effect of perfusion with KPM buffer. The time after perfusion in each frame is the same as in (A). This fiber lost 49% of its OD but only 10% of its length during the interval shown. Numerals in each frame indicate elapsed time from perfusion (in min) in (A) and (B). Bar, 5 μm. (C) Mean cumulative loss of length over time of 32 fibers perfused with S100 plus YOP (▲) compared with that of 18 fibers perfused with KPM buffer alone (⚫). (D) Cumulative loss of OD of the same fibers analyzed in (C) (△, S100 + YOP in KPM; ◯, KPM).

YOP-induced retraction in vitro closely paralleled retraction in whole sperm. In crawling sperm, localized cytoskeletal disassembly at the base of the lamellipod pulls the cell body forward (Fig. 2A, movie S2) (13). When a bead was attached to the end of fiber opposite the vesicle, retraction pulled the attached bead forward toward the vesicle (Fig. 2B, movie S3) in the same way that the disassembling fiber complexes pull the cell body along as the cell crawls (17). When protrusion was blocked by treatment of sperm with acetate buffer at pH 6.35, the fiber complexes detached from the lamellipod membrane, and the tension generated at the base of the lamellipod pulled the fiber complexes rearward while the cell body remained stationary (Fig. 2C) (13). We observed the same pattern in vitro when the bead was anchored to the substratum: Retraction pulled the vesicle-bearing end of the attached fiber toward the bead (Fig. 2D). Although retraction in vivo occurs at the acidic end (pH 6.5) of a lamellipodial pH gradient in crawling sperm and in a uniformly acidic lamellipod (pH 6.35) in acetate-treated cells (13, 18), treatment of fibers in vitro with acetate buffers at pH 6.35 to 6.5 did not induce retraction and neither low pH nor a pH gradient was required for YOP-induced retraction. These observations suggest that pH does not trigger cytoskeletal retraction in vivo but, instead, may contribute indirectly, perhaps by activating an endogenous phosphatase.

Fig. 2.

Comparison of patterns of retraction in vivo and in vitro. (A) Differential interference contrast (DIC) micrograph of a crawling Ascaris sperm. The boxed area is enlarged and shown as a time-lapse sequence (5-s interval between frames) to illustrate cytoskeletal dynamics and cell body retraction. Assembly of the fiber complexes results in protrusion of the leading edge. The arrowhead indicates a branch in a fiber complex that remains stationary with respect to the substratum but treadmills through the lamellipod as the cytoskeleton is taken apart at the base of the lamellipod where the cell body (cb) is pulled forward (retracts). (B) Retraction of a fiber with a bead attached at the end opposite the vesicle. As the fiber shortens it moves the attached bead in the same way that the MSP cytoskeleton pulls the cell body forward in crawling sperm. (C) DIC micrograph of an intact sperm treated with HKB buffer containing 20 mM sodium acetate (pH 6.35). The time-lapse sequence of the enlarged boxed area shows that lowering cytoplasmic pH stops cytoskeletal assembly and protrusion of the leading edge and detaches the cytoskeleton from the lamellipod membrane. Cytoskeletal disassembly continues and the fiber complexes are pulled progressively rearward through the lamellipod toward the cell body. (D) Retraction of a fiber with an attached bead when the bead is anchored to the substratum but the fiber is not produces a pattern like that observed in acid-treated sperm. The vesicle-bearing end of the fiber is pulled progressively toward the bead as the fiber shortens. Numerals in (B) and (D) indicate elapsed time (in min) from the start of retraction. Bars, 5 μm.

YOP appears to switch fibers from growth to retraction in the in vitro system by altering the tyrosine phosphorylation of key components. Consistent with this idea, sodium orthovanadate, a potent and broad-spectrum tyrosine phosphatase inhibitor, blocked the effects of YOP on fiber dynamics. Thus, fibers perfused with S100 containing both YOP and 1 mM orthovanadate not only failed to retract or lose OD but, if ATP was added, continued to grow (Fig. 3A). Fibers treated with YOP in assembly buffer without S100 stopped growing and lost OD, but shortened to a lesser extent and much more slowly than when S100 was present (Fig. 3B). Those treated with S100 without YOP also failed to retract and, if ATP was added, continued to grow (Fig. 3C). Previous work (19) showed that MSP polymerization at the vesicle surface is orchestrated by a 48-kD tyrosine-phosphorylated integral membrane protein (p48) and that YOP can stop fiber growth by dephosphorylating this protein. However, dephosphorylation of p48 cannot by itself account for the effect produced by YOP because, in the presence of S100, YOP not only stops fiber growth but also initiates fiber retraction, indicating that the activation of retraction by YOP requires one or more additional soluble components supplied by S100.

Fig. 3.

Retraction requires YOP plus an additional component in S100. The time after perfusion (in min) is indicated in each frame in (A) to (C). (A) Fiber perfused with S100 containing YOP, 1 mM sodium orthovanadate, and 1 mM ATP. The fiber continued to grow at its vesicle-associated end (arrow) after perfusion and did not shorten or lose OD, indicating that orthovanadate blocked the effects of YOP. (B) Fiber perfused with YOP in KPM assembly buffer. Without S100 present the fiber lost OD but exhibited very little shortening. (C) A similar sequence showing a fiber perfused with S100 containing 1 mM ATP but without added YOP. The fiber continued to grow at its vesicle-bearing end and did not retract. (D) Rescue of a retracting fiber. The left panel was obtained 10 s after perfusion of a growing fiber with S100 plus YOP, which initiated fiber retraction (second panel). The third panel shows the fiber 10 s after washing out the retraction-inducing solution with S100 containing 1 mM ATP but lacking YOP. Retraction stopped and fiber growth resumed, producing a new segment just behind the vesicle (arrow) with a higher OD than that of the retracted portion of the fiber. The retracted portion stopped shortening but did not regain its OD, whereas the new segment continued to lengthen (fourth panel) and had an OD comparable to that of the original fiber before YOP treatment. (E) EM of a platinum-shadowed retracted fiber induced to resume growing by perfusion with S100. The fiber diameter and filament density in the retracted segment (lower right) are reduced compared with that in the newly grown segment, but the filaments are arranged in a meshwork in both portions of the fiber. Bars, 5 μm.

Retracting fibers could be rescued (i.e., induced to resume growing) by replacement of S100 plus YOP with S100 containing ATP. Under these conditions, retraction stopped and fiber growth resumed at the vesicle surface (Fig. 3D). However, the retracted portion did not regain its OD, and so these fibers comprised two segments: the shortened remnant of the original fiber connected to a newly constructed segment generated by reactivation of vesicle-associated filament polymerization. Electron microscopy (EM) examination of these rescued fibers indicated that both the fiber diameter and the filament density in the retracted segment were reduced substantially compared with those of the newly assembled segment that had not undergone retraction (Fig. 3E). However, retraction did not change the general arrangement of filaments; in each portion of the fiber, the MSP filaments packed into a similar meshwork.

Reconstitution of both protrusion and retraction in vitro in the same motility apparatus provides a powerful experimental system for exploring how the forces for amoeboid cell motility are generated. Although the precise mechanism by which tyrosine phosphatase activity leads to retraction remains to be defined, it is unlikely that the enzyme activates an actomyosin-like contraction because MSP filaments lack the structural polarity required for the operation of motor proteins (20). Moreover, retraction does not require either ATP, the energy source for conventional motors, or the pH gradient present in the sperm lamellipod (18), which, in the absence of filament polarity, might provide directionality for a motor. Instead, our observations of fiber retraction in vitro imply that the mechanism of retraction in nematode sperm is essentially the converse of that of protrusion. For example, protrusion requires ATP and is blocked by YOP, whereas retraction is activated by YOP and is ATP-independent. Part of the ATP-derived energy used to assemble the cytoskeleton and generate protrusion may be stored in the fiber complexes and released to power retraction when the cytoskeleton is taken apart, consistent with a recent physical model of nematode sperm locomotion (21) and analogous to the way in which energy stored in the polymer lattice during microtubule polymerization can be released to produce movement associated with microtubule disassembly (22, 23). Depolymerization creates sparser filament density in the gel, but this alone cannot account for retraction because fibers perfused with buffer without YOP lost OD but did not contract (Fig. 1, B to D). Therefore, there must also be some rearrangement of the remaining filaments to generate shortening. Because gel behavior is sensitive to cross-linking (24), it may be that the filaments within a fiber are held by a cross-linking protein. YOP could break these cross-links and allow the gel to shrink or allow filaments to slide over one another to maximize interactions between filaments and so generate fiber shortening, analogous to some model nanotechnology systems (25).

Although amoeboid movement was originally thought to be powered primarily by an actomyosin-based sliding of filaments analogous to that seen in muscle (7), there is an emerging consensus that actin polymerization dynamics alone generate lamellipodial protrusion (2). Myosin has also frequently been proposed to be involved in retraction. Myosin II forms bipolar filaments associated with the actin cytoskeleton in the posterior portion of the lamellipod in cells such as fish epithelial keratocytes (9), although direct evidence linking myosin with retraction has not been obtained. Myosin II–null Dictyostelium cells are still able to move (26, 27), and recent work indicates that in these cells myosin II may be more important for detaching the rear of the cell from the substrate than in generating the retraction force that pulls the cell body forward (10).

Our results show that in the simple, stripped down MSP locomotory apparatus of Ascaris sperm, forward movement of the cell body is coupled to cytoskeletal retraction and that retraction can occur in vitro without the obvious involvement of motor proteins. Moreover, this system provides a platform in which to probe the precise mechanism of cytoskeletal retraction and how this process can be harnessed to pull the cell body.

The marked similarity in the patterns of cytoskeletal dynamics in nematode sperm and in several types of actin-rich crawling cells, in which filament networks assembled at the leading edge flux rearward and are taken apart deeper in the lamellipod (2, 5, 11), makes it plausible that a similar mechanism associated with cytoskeletal disassembly also contributes to cell body retraction in many actin-based crawling cells.

Supporting Online Material

Materials and Methods


Movies S1 to S3

References and Notes

Stay Connected to Science

Navigate This Article