REVIEW

William M. Bement � Anna M. Sokac

Craig A. Mandato

Four-dimensional imaging of cytoskeletal dynamics in Xenopus oocytesand eggs

Accepted October 30, 2003

Abstract The Xenopus laevis (African clawed frog)system has long been popular for studies of bothdevelopmental and cell biology, based on a variety of itsintrinsic features including the large size of Xenopusoocytes, eggs, and embryos, and the relative ease ofmanipulation. Unfortunately, the large size has alsobeen considered a serious impediment for high-resolu-tion light microscopy, as has the heavy pigmentation.However, the recent development and exploitation of4D imaging approaches, and the fact that much of whatis of most interest to cell and developmental biologiststakes place near the cell surface, indicates that suchconcerns are no longer valid. Consequently, theXenopus system in many respects is now as good asother model systems considered to be ideal formicroscopy-based studies. Here, 4D imaging and itsrecent applications to cytoskeletal imaging in Xenopusoocytes and eggs are discussed.

Key words 4D imaging � cytoskeleton � Xenopus �eggs � embryos

Introduction

Of all the sub-disciplines of biology, cell biology hasbeen most consistently defined and advanced based onthe union of different experimental approaches. Indeed,the birth of cell biology is generally dated to the happymarriage of light and electron microscopy with bio-chemistry. Further, the subsequent incorporation ofmolecular biology (i.e., recombinant DNA-based ap-proaches) and genetics into the repertoire of cellbiologists has led to advances in our understanding ofcell structure, dynamics, and function that can only bedescribed as astonishing. The same is true for develop-mental biology, which has benefited enormously fromthe integration of multiple disciplines.

Consequently, the rise, decline, and resurrection ofspecific cell and developmental model systems has beenclosely tied to their real or perceived utility with respect todifferent approaches. For example, echinoderm eggs andembryos, long favored for developmental biology, werealso extremely popular for studies of the cytoskeleton andcell cycle during the 1970s and 1980s based on their utilityfor both biochemistry and microscopy. However, thesesystems have since been largely supplanted by organismssuch as yeast and flies, which allow genetic approachesand are more amenable to molecular biology-basedmanipulations (for a variety of reasons, expression ofexogenous constructs is not as straightforward inechinoderm eggs as in other systems).

Xenopus laevis (African clawed frog) oocytes, eggs,and embryos have likewise undergone something of adecline in popularity in the last decade. While ideal forexpression of exogenous proteins and mRNAs, andoutstanding for biochemical approaches including

William M. Bement ( .*)Department of ZoologyProgram in Cellular and Molecular BiologyUniversity of Wisconsin, Madison1117 West Johnson StreetMadison, WI 53706, USATel: (608) 262-5683Fax: (608) 262-9083e-mail: wmbement@wisc.edu

Anna M. SokacDepartment of Molecular BiologyPrinceton UniversityWashington RoadPrinceton, NJ 08544, USA

Craig A. MandatoDepartment of Anatomy and Cell BiologyFaculty of MedicineMcGill University3640 University StreetMontreal, Quebec H3A 2B2, Canada

U.S. Copyright Clearance Center Code Statement: 0301–4681/2003/7109–518 $ 15.00/0

Differentiation (2003) 71:518–527 r International Society of Differentiation 2003

generation of cell-free extracts, they suffer from both areal lack of genetic attack (but see Discussion) and aperceived lack of microscopy-based approaches.

It is the latter point that this review seeks to address.Specifically, recent advances in imaging approacheshave revealed that Xenopus oocytes and eggs are in factappropriate subjects for high-resolution, live cellimaging of the cell cortex, which has long beenconsidered a critical nexus for both cell and develop-mental processes (Sardet et al., 2002). We will considerrecent studies of cytoskeletal dynamics in Xenopusoocytes and eggs and, briefly, several other features ofthe system that have facilitated the imaging approachand hold great promise for future studies.

The material is presented in as simple a manner as weare able. We have made no effort to cover the theory thatunderpins confocal or 4D (four-dimensional) imagingbecause other reviews that are more than adequate areavailable (e.g., Pawley, 1995; Thomas et al., 1996). At theother end of the spectrum, we have avoided specificdetails about how particular experiments are done simplybecause these will depend on the imaging system aparticular investigator has and the type of biologicalprocess they wish to study. The goal is to illustrate, in aspainless a manner as possible, the tremendous, but as yetuntapped potential of Xenopus oocytes and eggs forimaging of cell and developmental biology-based pro-cesses. Our hope is that a simple description of 4Dapproaches and some of the general pitfalls likely to beencountered while using them will prompt others to trytheir hands with fluorescent, live cell imaging.

While we will not consider them here, many of theapproaches and considerations presented below are alsorelevant to cell-free extracts derived from Xenopus eggs.Such extracts have long been employed for analysis ofbasic cell biological processes including cell cycleregulation (Lohka and Masui, 1983; Murray andKirschner, 1989), nuclear envelope assembly (Newportet al., 1990), and even programmed cell death (Evanset al., 1997), to name a few. More recently, egg extractshave been used to demonstrate interactions between thedifferent cytoskeletal systems (Sider et al., 1999; Water-man-Storer et al., 2000; Weber and Bement, 2002), andto reconstitute basic cortical processes (Marrari et al.,2003). As noted above for intact cell systems, there is noa priori reason why 4D imaging would not work forfractionated systems. Similarly, in principle, what worksfor the Xenopus system should also be applicable tooocytes, eggs, and embryos from marine organisms,mammals, flies, and other classic models.

Methods and Results

Four-dimensional imaging

Two features of Xenopus oocytes, eggs, and embryoshave long promoted the assumption that they are poor

candidates for imaging. First, their immense size( � 1.2mm diameter) implies that most of their volumewill be inaccessible to the light microscope, due to bothout-of-focus information and spherical aberration.Second, full grown oocytes, eggs, and embryos have anumber of compartments including pigment granulesand yolk platelets, which diffract light and renderimaging of the deep cytoplasm extremely difficult,particularly in living samples. Because F-actin, micro-tubules, and intermediate filaments are extended proteinpolymers, the above factors are considered especiallyproblematic for studies of the cytoskeleton, in that agiven polymer system may occupy several focal planes.Thus, while earlier confocal reconstructions of fixedsamples have permitted detailed analyses of the oocyteand egg microtubule (Gard, 1991), actin (Roeder andGard, 1994), and cytokeratin (Gard et al., 1997)networks, in only a handful of cases have live oocyteor egg studies been assayed (Clarke and Allan, 2003).

However, in the last decade, it has become apparentthat an astonishing amount of critical informationabout cytoskeletal dynamics can be extracted fromliving Xenopus oocytes and eggs. For example, Gardwas able to follow meiotic spindle rotation in livingXenopus eggs using time-lapse, confocal microscopy(Gard, 1992). Using the same approach, Taunton et al.(2000) followed actin dynamics in living Xenopus eggs,and provided the first demonstration that actin ‘‘co-mets,’’ previously shown to result from a variety ofexperimental interventions in different cell types (Ro-zelle et al., 2000), was a natural event of egg activation.

We have applied 4D confocal imaging to the analysisof F-actin during cortical flow (Benink et al., 2000),actomyosin (Mandato and Bement, 2001), and micro-tubule (Mandato and Bement, 2003) dynamics duringwound healing in Xenopus oocytes, and analysis of actindynamics during egg activation (Sokac et al., 2003).Four-dimensional imaging simply means the acquisitionof a stack of images (in other words, multiple imagescollected at different focal planes within a given field)for each time point in a time-lapse series (Fig. 1). Thus,at each time point, the focus motor moves the objectivelens (and thus the focal plane) a defined number of stepsof equal distance through the sample, and images arecollected at each focal plane (this is also known asoptical sectioning). Once all of the images in a stack arecollected, the objective is repositioned by the focusmotor to the starting focal plane in preparation for thenext time point, and the process repeats itself. Theimages for each time point are then saved as digitalimage stacks for later reconstruction.

This approach permits the investigator to captureinformation in more than one focal plane in livingsamples with a minimum of out-of-focus information.This is particularly important for large cells such asoocytes and eggs (and, of course, embryos) and forimaging of structures that typically occupy more than

519

one focal plane. An example is provided in Fig. 2. Thetwo rows of images were collected from the samesample. In the top row of images, only a single focalplane per time point is shown; in the bottom row, astack corresponding to 13 focal planes per time point isshown. The series shows the assembly of F-actin‘‘coats’’ around exocytosing cortical granules (see Sokacet al., 2003). Briefly, when exocytosis is triggered,extracellular fluorescent dextran (red) is incorporatedinto the compartments created by fusion of the cortical

granules with the plasma membrane. Actin (green)subsequently assembles around these compartments.

Even casual inspection is sufficient to reveal strikingdifferences between the two sets of images. In the singleoptical plane images, only a subset of the total numberof cortical granules that exocytose are clearly observed,because others are mostly above the chosen focal plane.In contrast, the series derived from 4D imaging showsall of the cortical granules that exocytose in this field ofview. Further, because this 4D data set was rendered as

Fig. 1 Schematic diagram showingthe basic principle of 4D imaging.The top part of the image shows aXenopus oocyte mounted for ima-ging by the objective lens. Thesmall arrows to the right indicatemovements of the objectivethrough the sample to generateoptical sections. The large arrowsindicate the return of the objectiveto the original focal plane. T1, T2,etc. indicate the individual timepoints. The bottom part of theimage shows representations of theinformation obtained from eachtime point (raw 4D data set) andhow it can be subsequently pro-cessed: compression into projec-tions (2D 1 time) or rendering intovolumes (3D1time).

Fig. 2 Confocal series showing single optical planes (top row)or rendered volumes (bottom row) from the same time-lapse seriesof actin coat assembly during Xenopus egg cortical granuleexocytosis. Actin (green) encircles and closes around exocytosingcortical granules, which are revealed by their incorporationof fluorescent dextran (red). Note that in the top row, only

two exocytosing cortical granules are easily seen (arrowheads) while in the bottom row, several others are quiteobvious (arrows). In addition, the bottom row provides theimpression of depth, such that the actin coats appear to besurrounding the exocytosing cortical granules and closing over theirtops.

520

a volume (see below), depth information is provided.That is, the actin coats appear to be extending upwardover and eventually engulfing the compartments createdby the exocytosing cortical granules.

2D+time versus 3D1time: Once a 4D data set iscollected, there are two major ways of reconstructing it.Specifically, after acquisition, the image stacks fromeach time point can either be compressed into a single2D representation for each time point or ‘‘rendered’’into volumes (3D representations) for each time point(Fig. 1). The former approach is appropriate forsituations in which information from the X and Ydimensions of the field of view greatly exceeds that fromthe Z dimension (Fig. 3). The latter is appropriate forsituations in which information from the Z dimension isroughly equal to or exceeds that from the X and Ydimensions (Fig. 2). Thus, if the process in question isone in which most of the action is represented by lateralmovement, compression into a 2D representation ateach time point is indicated, while if the process inquestion is one in which much of the importantmovement is directed inward (i.e., in the Z plane),rendering is indicated.

Obviously, these distinctions represent ends of acontinuum. In an ideal world, 4D data sets wouldalways be rendered as volumes, because that way noinformation would be lost and there is nothing to stopthe investigator from later presenting them as 2D1timeif desired. However, there are three costs to renderingvolumes—storage space, processing time, and softwarepurchase. With respect to storage space, if a given raw4D data set is converted to 2D+ time prior to saving,the file size is reduced by as many times as there arefocal planes in the original image stack. For example, a4D data set derived from an experiment in which tenoptical sections were collected for each time point willbe ten times bigger if stored raw (as needed for volumerendering) versus compressed into a 2D representation.In addition, further space is required to store therendered volumes, at a minimum the same amount asrequired for the original raw file, but more frequentlymuch more, as movies and still images are amassed fordifferent viewing angles, crops, and so on. With respectto time, rendering volumes from 4D data sets takesanywhere from 5 to 30min, depending on the size of theset in question. This is in addition to the amount of timerequired to convert rendered volume sequences intomovies. In contrast, 2D+ time data sets can beimmediately converted into movies or other outputssuitable for publication. With respect to softwarerequired for rendering volumes for viewing of 4Dmovies, at present, it cannot be done without thepurchase of relatively sophisticated software packages(see below).

Thus, if a given sample is sufficiently flat, it may bemore desirable to save it as a 2D1time series. This doesnot mean, however, that 4D image collection has not

Fig. 3 Confocal series showing Xenopus oocyte wound healing. Theoriginal 4D data set has been collapsed into 2D 1 time. Actinfilaments (red) and microtubules (green) move toward the wound(arrows; asterisk marks a single microtubule-actin filament com-plex) over time. The area of movement in X–Y is quite extensive,extending from the wound edge (W) to the area indicated by thelarge arrowheads. The use of 4D collection makes it quite apparentthat the microtubules and actin filament cables are extensivelyaligned (small arrowheads).

521

made an important difference. For example, woundingof Xenopus oocytes results in the rapid assembly andaccumulation of actin, myosin-2, and microtubulesaround wound borders (Mandato and Bement, 2001,2003). The wound-associated actin and microtubulesare concentrated within � 2–10 mm of the plasmamembrane (Z information), while the wound and thearea around it in which actin and microtubulesreorganize may be 100 mm or more away from thewound (X–Y information). Further, microtubules andcables of actin filaments move toward the wound withtheir long axes perpendicular to the wound border. Inthis case, compression of stacks from each time pointinto a 2D representation is appropriate (Fig. 3).However, were the imaging done in a single opticalplane, only a fraction of the moving microtubules andactin filament cables would be in focus at each timepoint. As a result, not only would some cytoskeletalelements fail to appear at all (as above with theexocytosing cortical granules), others would appearand disappear from time point to time point as theymove between focal planes, making it difficult orimpossible to track them for more than a frame or two.

By way of counter-example, consider the exocytosisof cortical granules presented above. Cortical granules

are � 1–3 mm in diameter, and occupy a space within0.1–5 mm of the plasma membrane. For reasons relatedto speed (see below), when imaging cortical granuleexocytosis, the X–Y dimensions range from � 20 � 20to � 5 � 5 mm. Actin coats assemble around theexocytosing cortical granules, moving both upwardand downward in Z. Thus, in this case, rendering theoriginal 4D data set into volumes (3D1time) is notonly appropriate, it is essential because much ofthe important information is represented by movementin Z.

Perspective: When rendered as volumes, one of themost powerful features of 4D image sets is the fact thatthey provide one with the opportunity to view thesample from an unlimited number of perspectives bysimply rotating the volume. The advantage of this isillustrated in Figs 4 and 5, which offer differentviewpoints of assembling actin coats. In Fig. 4, the‘‘apical’’ view of exocytosing cortical granules isequivalent to being outside the cell and looking downon the cell surface. The ‘‘basal’’ view is equivalent tobeing inside the cell, looking up at the underside of theplasma membrane. While the overall impression fromthe two views is similar, there are also obviousdifferences. For example, the basal view shows that

Fig. 4 Three different perspectives of the same time pointin a rendered volume depicting actin coat assembly. Whilethe apical and basal views at first appear quite similar, closerinspection shows that two of the exocytosing cortical granules(red; 1 and 2) are open on their exoplasmic surface (i.e., the surface

that faces outside the cell) but largely covered with actin(green) on their basal surface (i.e., the surface that faces thecytoplasm). The tilt view reveals the actin is not uniformlycoating the cytoplasmic side of one of the exocytosing corticalgranules.

Fig. 5 A time-lapse, Z perspective derived from volumes renderedfrom a 4D series of actin coat formation. Actin coats (green)

rapidly surround the exocytosing cortical granules (red) and thencompress upward.

522

the actin coats have completely covered the cytoplasmicsurfaces of three of the exocytosing cortical granuleswhile the apical view shows that the exoplasmic side ofthese same compartments are not yet closed off by actin.The tilt view provides yet another perspective, andreveals that the sides of some of the cortical granulemembranes are not completely covered with actin.

Importantly, all of these views were generated fromthe same volume. Further, while these represent a singletime point, with the full 3D 1 time series, theinvestigator can observe the dynamic process of actincoat assembly from any of these angles or, for thatmatter, from any other angle that seems appropriate.For example, in Fig. 5, a Z time series of actin coatassembly on exocytosing cortical granules is presented.It shows quite clearly that the actin coats compressupward over time, as well as giving a comparativelyprecise view of just how far into the cells the actin coatsextend. The possibilities here are, literally, infinite, aslong as the 4D data set is rendered into volumes. Inaddition to those viewpoints presented above, it is alsoeasy to generate cutaway views, thin slices, grazingsections, and so forth.

Hardware and software: One of the more surprisingaspects of 4D imaging is that it can be done with‘‘standard’’ imaging set ups. That is, any confocalmicroscope with software that allows collection of time-lapse Z series is capable of producing 4D data sets thatcan then be converted into either 2D1time or 3D1timeseries. For example, the data sets used to generate theimages shown above and in our published work on livecell imaging (Mandato and Bement, 2001, 2003; Sokacet al., 2003) were collected with a Bio-Rad 1024 laserscanning confocal microscope (Hercules, CA) mountedon a Zeiss Axiovert base (Thornwood, NY). The Bio-Rad LaserSharp software that comes with this systemas installed, allows collection of 4D data sets. It shouldbe pointed out that this is not something that is eithernew or specially ordered. In fact, the software inquestion is comparatively old and runs on MicrosoftOS2. Further, equivalent software for 4D data setcollection is now considered a standard feature for mostnew confocal imaging systems, in spite of the fact thatmost investigators still do not use it.

Once the raw 4D data sets are collected, they aretypically transferred to a desktop computer for analysisand rendering. Because a typical 4D data set may be 50–500MB depending on the experiment in question, a lotof memory is required both for the instrument used forimage collection, and for that used for analysis andrendering. However, because the typical upper-enddesktop computer has 60–80GB or more, this require-ment is not hard to satisfy.

If the 4D data sets are to be immediately converted to2D1time, this can be accomplished by LaserSharp orequivalent software that comes with the imagingsystem. They can then be viewed and manipulated

using two free programs—Object Image and Image J,both of which are available on the web. If the 4D datasets are to be rendered, we recommend Volocity, a 4Dimaging package from Improvision (Lexington, MA).We have used Volocity for 2 years and, at the risk ofsounding unduly partisan, have not seen anythingcomparable from other vendors. It works with PCsand Macs, recognizes a variety of file types, and has auser-friendly interface. More importantly, it producestrue 3D1time series that can be rotated during viewing,which is absolutely essential when trying to characterizecomplex, dynamic processes that have a significantdepth. Once the desired orientation is determined,individual frames or whole movies can be saved inone of several standard formats (e.g., Tiff, Quicktime).

Quantification

Biological imaging is becoming increasingly quantita-tive. Importantly, 4D imaging facilitates quantificationin at least three ways. First, by collecting all of therelevant information rather than just what is availablefrom a single focal plane, it allows the investigator toavoid underestimation of numbers of objects under-going a particular process. Second, it allows theinvestigator to quantify rates of movement that occurin Z. Third, it also allows the investigator to quantifyother parameters in Z, such as changes in fluorescenceintensity over time. Obviously, for Z quantifications,volume rendering is essential, as compression of the raw4D data sets into 2D 1 time loses all Z information.

Probes

In keeping with its broad applicability, the samefluorescent probes used for other live imaging applica-tions work just as well for 4D imaging. We have used avariety of commercially available probes for 4Dimaging in Xenopus oocytes and eggs, as well as anumber of eGFP (enhanced green fluorescent protein)fusions generated in our lab. With the exception offluorescent lipid markers, fluorescent probes are usuallyintroduced into Xenopus oocytes, eggs, and embryos bymicroinjection. In contrast to cultured cells, or evenmany other egg/embryo systems, microinjection in theXenopus system is remarkably easy. We routinelydouble or triple inject oocytes, and injecting severalhundred oocytes in the space of an hour or two is not aserious challenge, even for those of us (e.g., W.M.B.)who have been known to have difficulty in zipping upour trousers. As described below, this ease of injectiongreatly facilitates systematic characterization of newprobes.

Potential dominant-negative effects: Because there isno such thing as a probe that does not have some kindof dominant-negative effect when used at high concen-trations, it is strongly recommended that each new

523

probe be carefully tested for its effects on the process ofinterest. For example, we have yet to find a fluorescentlipid probe that does not inhibit actin coat assembly(Sokac et al., 2003). Because this includes R18 and FM-143, probes that have been successfully used to followexo-/endocytosis in other systems (Terasaki, 1995;Whalley et al., 1995), it follows that it cannot beassumed that what works in one system or process willnecessarily work in similar processes or systems. Wehave also observed less extreme phenotypes fromfluorescent actins, but even these at high concentrationscan significantly affect actin assembly itself. In addition,most probes derived from GFP fused to domains ofproteins with specific binding partners (e.g., eGFP fusedto the PH domain of phospholipase C delta, whichbinds to PIP2) have at least two potential undesirableeffects if used at high concentrations: (1) binding toalternative targets and (2) suppression of pathwayscontrolled by the target of interest. Thus, when startingwith a new probe, we typically inject it at three to fivedifferent concentrations in a single batch of oocytes,and then image, with an eye toward finding the lowestconcentration that provides strong signal at lowimaging power (preferably 3% laser power or less)without disrupting the process of interest. The ability toinject multiple cells quickly and with more than oneprobe has greatly facilitated our progress in imaging. Itis difficult to conceive of doing the same experiments in,say, cultured mammalian cells, either by injection ortransfection.

Proteins and RNAs: Injection of fluorescent proteinprobes is often more convenient than injecting mRNAbecause they are ready for imaging just as soon as theyhave diffused throughout the cell, whereas mRNA hasto express for several hours to overnight. In addition,injection of a constant amount of fluorescent proteinproduces a fairly constant signal, whereas the signalobtained after injection of an mRNA encoding eGFPfusions will vary somewhat depending on the rate ofprotein synthesis, which differs between differentbatches of oocytes. On the other hand, it is far easierto generate mRNA encoding eGFP fused to specificproteins than it is to purify and fluorescently label mostproteins. Further, once made, mRNA requires nospecial preparation prior to injection, whereas fluor-escent proteins typically must be centrifuged at100,000 g for 15–30min before injection. This is not tosay that the eGFP fusion approach always works,because it doesn’t. On the other hand, it works oftenenough to make it worth trying for each new protein orlipid probe of interest. The success rate goes up if oneconsistently tries both N- and C-terminal fusions.

In contrast, we have had virtually no success using acommercially available red fluorescent protein, presum-ably because it is tetrameric and very slow to express.Hopefully, the recently generated mutant forms of redfluorescent protein (RFP) that are monomeric (Camp-

bell et al., 2002) and comparatively fast to express(Bevis and Glick, 2002) will function as well as eGFP.Thus, most of our double-label 4D imaging has beendone using eGFP fusions together with proteins orother probes directly labeled with red-emitting fluor-ophores. We have not yet tried other fluorescent proteinvariants (e.g., YFP or CFP) but these should also work,assuming one has the appropriate filter sets.

Other probes: Several fluorescent toxins are avail-able, and have proven quite useful for 4D imaging ofthe cytoskeleton. Fluorescent phalloidins labeled withseveral different fluorophores can be purchased from avariety of sources and are useful because they bindspecifically to actin filaments (as opposed to actinmonomer). When used alone with 4D imaging, theyprovide an excellent marker for contraction-basedmovement of actin filaments (Mandato and Bement,2001), and when used in conjunction with fluorescentactin, they can help the investigator distinguish betweenstable and dynamic (i.e., rapidly assembling anddisassembling) actin (Mandato and Bement, 2001;Sokac et al., 2003).

Fluorescent taxol (aka paclitaxel) is also soldcommercially (by Molecular Probes, Eugene, OR).Taxol is analogous to phalloidin, in that it binds tomicrotubules but not disassembled tubulin dimer.Fluorescent taxol works well as a label for stablemicrotubules when used in conjunction with 4Dimaging in Xenopus oocytes (Mandato and Bement,2003). It can also be used with fluorescent phalloidin topermit simultaneous visualization of F-actin andmicrotubules (Fig. 3; see also Mandato and Bement,2003).

Both phalloidin and taxol must be used with care,because both of them stabilize (i.e., prevent thedisassembly of) the polymers they interact with. Thatis, phalloidin stabilizes actin filaments and taxolstabilizes microtubules. Thus, when used at highconcentrations, they result in striking increases in levelsof actin filaments and microtubules in the cell,respectively. Such effects are not just a serious problemfor those interested in studying cytoskeletal dynamics,they also have a widespread impact on virtually everyother aspect of cellular and developmental function.Consequently, as above for protein and mRNA probes,it is best to carefully titer the amounts of these probesinjected.

General experimental concerns

Experimental design will vary sharply depending on thesystem being employed, the process being studied, andthe type of instrumentation used. Nevertheless, thereare several issues that are likely to arise in manydifferent experimental situations. For example, whileone is naturally tempted to try the most complicatedexperiment first, in the long run it is better to proceed in

524

a stepwise fashion. Thus, once a given probe has beenvetted, it is generally a good idea to first use singleoptical plane imaging over time to characterize tempor-al aspects of its distribution as well as optimal imagingsettings (laser power, etc.).

If two different fluorophores are being employed, itmay be necessary to modulate the amount of eachinjected, to avoid problems resulting from synergisticdominant-negative effects. Another common problem is‘‘bleed-through’’ of signal from one channel to theother. Even very good filter sets can permit bleed-through if the signal from one fluorophore is too high,so it is essential to confirm in double-label experimentsthat signal from one channel is not contributing to theinformation collected in the other channel. Because it iseasier to image in a single optical plane over time than itis to collect 4D image sets, one or two pilot experimentsto address potential bleed-through problems can savemuch aggravation (or embarrassment) down the road.

The speed with which images are collected is alsosomething that must be tailored to each experimentalsituation. The speed will depend on some factors thatcannot be controlled, such as the rate of focus motormovement. However, there are other factors that can becontrolled, including the number of focal planesimaged, the size of the area being imaged (in X–Y),the speed of the laser scan (if a laser scanning confocalmicroscope is used), the number of scans, and theinterval of sampling. The sample and process beingstudied will tend to dictate the number of focal planesand the interval of sampling, which means that theinvestigator may have to change the other parametersto achieve satisfactory imaging. For example, forimaging wounds at low magnification, we often use a512 � 512 or even 1024 � 1024 pixel X–Y box size,because only four to six focal planes need to be collectedfor each time point and the time points may be as muchas 30 s apart. In contrast, to image actin coat assembly,as many as 15 focal planes may be necessary, andsampling intervals need to be 5–10 s. Consequently, weuse a much smaller box size (256 � 256 or even128 � 128 pixels).

Four-dimensional imaging can be sensitive to samplemovement during imaging, if the movement, either inX–Y or Z, is significantly faster than all of the focalplanes in a single time point can be collected. As aconsequence, manipulations that destabilize the cellsurface or result in, say, cell contraction, can beproblematic for imaging at high magnification, becauseincreasing magnification progressively exaggerates theeffects of movement. For example, Xenopus egg activa-tion triggers cortical contraction, in which the entire cellcortex is displaced toward the animal pole. If allowed tooccur unchecked, this makes it difficult to image actincoat assembly, which occurs on a time frame thatoverlaps with cortical contraction. To limit thisproblem, we include lectins in the imaging medium,

which stabilize the cell surface and reduce contractionby cross-linking cell surface proteins (Sokac et al.,2003). However, even lectin addition cannot preventrapid cell surface movements following disruption of F-actin followed by induction of exocytosis. The onlycurrent remedy for this problem is patience andrepetition.

Discussion

Four-dimensional microscopy is not simply a way toproduce pretty images. Indeed, it has been essential forboth our interpretation of processes dependent on thecortical cytoskeleton, as well as providing mechanisticinformation. For oocyte wound healing, while it wasclear from confocal analysis of fixed samples that actinand myosin-2 accumulate around wound borders(Bement et al., 1999), it was not at all obvious howthey got there. Four-dimensional analysis revealed,surprisingly, that actin and myosin-2 are recruited towound edges as a result of both local assembly and flow(translocation) through the cortex (Mandato andBement, 2001). In addition, 4D imaging was alsoessential for revealing apparent physical interactionsbetween translocating F-actin and translocating micro-tubules, as well as showing that microtubules (like actinand myosin-2) are also assembled near wound borders(Mandato and Bement, 2003). Based on these findings,we proposed that cytoskeletal reorganization aroundwounds results both from local signals that controlassembly of actin, myosin-2, and tubulin, and fromphysical interactions between these systems that pull F-actin and microtubules to the wound border viamyosin-powered contraction.

For actin coat assembly, while single optical plane,time-lapse imaging revealed that assembling actinassociated with exocytosing cortical granules, it didnot show much more. In contrast, 4D imaging showedthat actin assembly commenced from the point wherethe cortical granules fused with the plasma membraneand that actin coats appear to push inward, compres-sing the compartment created by the exocytosingcortical granule (Sokac et al., 2003). Based on theformer finding, we proposed that mixing of componentsof the plasma membrane and the cortical granulemembrane acts as the trigger for actin coat assembly.Based on the latter finding, we proposed that actinassembly itself provides the force for actin coatcompression.

Thus, 4D microscopy is a powerful tool for analysisof cytoskeletal dynamics in Xenopus oocytes and eggs.And this represents just a fraction of the potential forthis system, particularly considering the ease with whichexogenous constructs can be injected and expressed.For cell biology applications, analysis of spindle

525

dynamics (Becker et al., 2003), cytokinesis (Danilchiket al., 2003), receptor-mediated endocytosis of vitello-genin (Fagotto and Maxfield, 1994), and constitutiveexocytosis (Schmalzing et al., 1995) are just a samplingof obvious choices. For developmental biology, analysisof cortical RNA trafficking (Bubunenko et al., 2002;Zearfoss et al., 2003), cortical rotation (Marrari et al.,2000, 2003), embryonic wound healing (Bementet al., 1999; Kofron et al., 2002), gastrulation (Sundar-am et al., 2003), and neural tube formation(Barreto et al., 2003) should all be accessible to 4Dimaging.

When combined with other approaches, 4D imagingis even more powerful. For example, we have used it inconjunction with dominant-negative expression toassess the role of Xenopus myosin-1C during exo-/endocytosis (Sokac and Bement, 2002). It could also beemployed in conjunction with other reverse geneticapproaches in Xenopus, such as antibody injection(Becker et al., 2003), oligonucleotide-mediated RNAdegradation (Vernos et al., 1995), morpholinos (Heas-man et al., 2000), and siRNA (Anantharam et al.,2003). The ever-increasing number of sequences in theXenopus database will make such strategies increasinglyattractive.

Further, while forward genetic approaches in Xeno-pus laevis are impractical, the recent development ofXenopus tropicalis as a genetic model system (reviewedin Hirsch et al., 2002) suggests that in the future it willbe possible to combine all of the major tools in therepertoire of the modern cell and developmentalbiologist in a single system. That is, most of the otherfavored genetic models for metazoan cell and develop-mental biology are not particularly good for biochem-istry. In contrast, a single Xenopus oocyte, egg, orembryo provides sufficient material for a lane on anSDS gel (Ferrell and Machleder, 1998). Given the easeof microinjection and the facility with which exogenousconstructs can be expressed, it seems likely that it willnot be long before the Xenopus system is employed forstudies in which genetics, molecular biology, biochem-istry, and imaging are applied in combination to singleoocytes, eggs, and embryos.

Acknowledgments We are, as usual, grateful to the other membersof our labs for support. This work was supported by NationalInstitutes of Health award GM52932-04A1 to W.M.B. TheNational Science Foundation supported the acquisition of themicroscope used for the studies described via NSF9724515 toJ. Pawley.

References

Anantharam, A., Lewis, A., Panaghie, G., Gordon, E., McCrossan,Z.A., Lerner, D.J. and Abbott, G.W. (2003) RNA interferencereveals that endogenous Xenopus MinK-related peptides governmammalian K1 channel function in oocyte expression studies. JBiol Chem 278:11739–11745.

Barreto, G., Reintsch, W., Kaufmann, C. and Dreyer, C. (2003)The function of Xenopus germ cell nuclear factor (xGCNF) inmorphogenetic movements during neurulation. Dev Biol257:329–342.

Becker, B.E., Romney, S.J. and Gard D, L. (2003) XMAP215,XKCM1, NuMA, and cytoplasmic dynein are required for theassembly and organization of the transient microtubule arrayduring the maturation of Xenopus oocytes. Dev Biol 261:488–505.

Bement, W.M., Mandato, C.A. and Kirsch, M.N. (1999) Wound-induced assembly and closure of an actomyosin purse string inXenopus oocytes. Curr Biol 9:579–587.

Benink, H.A., Mandato, C.A. and Bement, W.M. (2000) Analysisof cortical flow models in vivo. Mol Biol Cell 11:2553–2563.

Bevis, B.J. and Glick, B.S. (2002) Rapidly maturing variants of theDiscosoma red fluorescent protein (DsRed). Nat Biotechnol20:83–87.

Bubunenko, M., Kress, T.L., Vempati, U.D., Mowry, K.L. andKing, M.L. (2002) A consensus RNA signal that directs germlayer determinants to the vegetal cortex of Xenopus oocytes. DevBiol 248:82–92.

Campbell, R.E., Tour, O., Palmer, A.E., Steinbach, P.A., Baird,G.S., Zacharias, D.A. and Tsien, R.Y. (2002) A monomeric redfluorescent protein. Proc Natl Acad Sci USA 99:7877–7882.

Clarke, E.J. and Allan, V.J. (2003) Cytokeratin intermediatefilament organisation and dynamics in the vegetal cortex ofliving Xenopus laevis oocytes and eggs. Cell Motil Cytoskeleton56:13–26.

Danilchik, M.V., Bedrick, S.D., Brown, E.E. and Ray, K. (2003)Furrow microtubules and localized exocytosis in cleavingXenopus laevis embryos. J Cell Sci 116:273–283.

Evans, E.K., Lu, W., Strum, S.L., Mayer, B.J. and Kornbluth, S.(1997) Crk is required for apoptosis in Xenopus egg extracts.EMBO J 16:230–241.

Fagotto, F. and Maxfield, F.R. (1994) Changes in yolk platelet pHduring Xenopus laevis development correlate with yolk utilizationA quantitative confocal microscopy study. J Cell Sci 107:3325–3337.

Ferrell, J.E. Jr. and Machleder, E.M. (1998) The biochemical basisof an all-or-none cell fate switch in Xenopus oocytes. Science280:895–898.

Gard, D.L. (1991) Organization, nucleation, and acetylation ofmicrotubules in Xenopus laevis oocytes: a study by confocalimmunofluorescence microscopy. Dev Biol 143:346–362.

Gard, D.L. (1992) Microtubule organization during maturation ofXenopus oocytes: assembly and rotation of the meiotic spindles.Dev Biol 151:516–530.

Gard, D.L., Cha, B.J. and King, E. (1997) The organization andanimal-vegetal asymmetry of cytokeratin filaments in stage VIXenopus oocytes is dependent upon F-actin and microtubules.Dev Biol 184:95–114.

Heasman, J., Kofron, M. and Wylie, C. (2000) b-catenin signalingactivity dissected in the early Xenopus embryo: a novel antisenseapproach. Dev Biol 222:124–134.

Hirsch, N., Zimmerman, L.B. and Grainger, R.M. (2002) Xenopus,the next generation: X. tropicalis genetics and genomics. DevDyn 225:422–433.

Kofron, M., Heasman, J., Lang, S.A. and Wylie, C.C. (2002)Plakoglobin is required for maintenance of the cortical actinskeleton in early Xenopus embryos and for cdc42-mediatedwound healing. J Cell Biol 158:695–708.

Lohka, M.J. and Masui, Y. (1983) Formation in vitro of spermpronuclei and mitotic chromosomes induced by amphibianooplasmic components. Science 220:719–721.

Mandato, C.A. and Bement, W.M. (2001) Contraction andpolymerization cooperate to assemble and close actomyosinrings around Xenopus oocyte wounds. J Cell Biol 154:785–797.

Mandato, C.A. and Bement, W.M. (2003) Actomyosin transportsmicrotubules and microtubules control actomyosin recruitmentduring Xenopus oocyte wound healing. Curr Biol 13:1096–1105.

526

Marrari, Y., Terasaki, M., Arrowsmith, V. and Houliston, E.(2000) Local inhibition of cortical rotation in Xenopus eggs by ananti-KRP antibody. Dev Biol 224:250–262.

Marrari, Y., Clarke, E.J., Rouviere, C. and Houliston, E. (2003)Analysis of microtubule movement on isolated Xenopus eggcortices provides evidence that the cortical rotation involvesdynein as well as kinesin related proteins and is regulated by localmicrotubule polymerisation. Dev Biol 257:55–70.

Murray, A.W. and Kirschner, M.W. (1989) Cyclin synthesis drivesthe early embryonic cell cycle. Nature 339:275–280.

Newport, J.W., Wilson, K.L. and Dunphy, W.G. (1990) A lamin-independent pathway for nuclear envelope assembly. J Cell Biol111:2247–2259.

Pawley, J.P. (1995) Handbook of biological confocal microscopy.2nd ed. Plenum Press, New York.

Roeder, A.D. and Gard, D.L. (1994) Confocal microscopyof F-actin distribution in Xenopus oocytes. Zygote 2:111–124.

Rozelle, A.L., Machesky, L.M., Yamamoto, M., Driessens, M.H.,Insall, R.H., Roth, M.G., Luby-Phelps, K., Marriott, G., Hall,A. and Yin, H.L. (2000) Phosphatidylinositol 4,5-bisphosphateinduces actin-based movement of raft-enriched vesicles throughWASP-Arp2/3. Curr Biol 10:311–320.

Sardet, C., Prodon, F., Dumollard, R., Chang, P. and Chenevert, J.(2002) Structure and function of the egg cortex from oogenesisthrough fertilization. Dev Biol 241:1–23.

Schmalzing, G., Richter, H.P., Hansen, A., Schwarz, W., Just, I.and Aktories, K. (1995) Involvement of the GTP binding proteinRho in constitutive endocytosis in Xenopus laevis oocytes. J CellBiol 130:1319–1332.

Sider, J.R., Mandato, C.A., Weber, K.L., Zandy, A.J., Beach, D.,Finst, R.J., Skoble, J. and Bement, W.M. (1999) Directobservation of microtubule-f-actin interaction in cell free lysates.J Cell Sci 112:1947–1956.

Sokac, A.M. and Bement, W.M. (2002) Xenopus myosin-1C isrecruited to cortical granule membranes upon egg activation andis required for cortical granule exocytosis. Mol Biol Cell 13:480a.

Sokac, A.M., Co, C., Taunton, J. and Bement, W. (2003) Cdc42-dependent actin polymerization during compensatory endocyto-sis in Xenopus eggs. Nat Cell Biol 5:727–732.

Sundaram, N., Tao, Q., Wylie, C. and Heasman, J. (2003) The roleof maternal CREB in early embryogenesis of Xenopus laevis. DevBiol 261:337–352.

Taunton, J., Rowning, B.A., Coughlin, M.L., Wu, M., Moon,R.T., Mitchison, T.J. and Larabell, C.A. (2000) Actin-dependentpropulsion of endosomes and lysosomes by recruitment of N-WASP. J Cell Biol 148:519–530.

Terasaki, M. (1995) Visualization of exocytosis during sea urchinegg fertilization using confocal microscopy. J Cell Sci 108:2293–2300.

Thomas, C., DeVries, P., Hardin, J. and White, J. (1996) Four-dimensional imaging: computer visualization of 3D movementsin living specimens. Science 273:603–607.

Vernos, I., Raats, J., Hirano, T., Heasman, J., Karsenti, E. andWylie, C. (1995) Xklp1, a chromosomal Xenopus kinesin-likeprotein essential for spindle organization and chromosomepositioning. Cell 81:117–127.

Waterman-Storer, C., Duey, D.Y., Weber, K.L., Keech, J.,Cheney, R.E., Salmon, E.D. and Bement, W.M. (2000) Micro-tubules remodel actomyosin networks in Xenopus egg extractsvia two mechanisms of F-actin transport. J Cell Biol 150:361–376.

Weber, K.L. and Bement, W.M. (2002) F-actin serves as a templatefor cytokeratin organization in cell free extracts. J Cell Sci115:1373–1382.

Whalley, T., Terasaki, M., Cho, M.S. and Vogel, S.S. (1995) Directmembrane retrieval into large vesicles after exocytosis in seaurchin eggs. J Cell Biol 131:1183–1192.

Zearfoss, N.R., Chan, A.P., Kloc, M., Allen, L.H. and Etkin, L.D.(2003) Identification of new Xlsirt family members in theXenopus laevis oocyte. Mech Dev 120:503–509.

527