Cytoarchitecture of size-excluding compartments in living cells

The common assumption that conditions inside a living cell can be extrapolated directly from the biochemistry of dilute, aqueous solutions is open to question. The interior of a cell is neither dilute nor homogeneous. Srere (1967) pointed out some time ago that the concentration of enzymes in the cell is orders of magnitude higher than concentrations used to study these enzymes in dilute solution. In fact, the total protein concentration in the cytoplasm may be as high as 200 to 300 mg/ml (Fulton, 1982; Lanni et al., 1985). Although it is not certain how much of this protein is truly soluble in the fluid phase, two models recently devised to account for experimental observations on the diffusion of fluorescent tracers in the cytoplasm of living cells suggest that the fluid phase may be as crowded as a 12 to 13% Ficoll or dextran solution (Hou et al., 1990; Kao et al., 1993). A large body of literature suggests that under such crowded conditions, the kinetics and/or regulation of biochemical pathways may bear little relation to what has been deduced from experiments in dilute solution (e.g. see Srere, 1967; Walsh and Knull, 1987; Jarvis et al., 1990; Minton, 1990). Rheologically, cytoplasm has properties more consistent with a visocoelastic gel than a homogeneous fluid (Taylor and Condeelis, 1979), and there is growing acceptance of the idea that both the nucleus and the cytoplasm are highly organized compartments in which a three-dimensional network of filaments forms a scaffold on which enzymes and nucleic acids are non-randomly supported (e.g. see Wolosewick and Porter, 1979; Clarke et al., 1985; Ben-Ze’ev, 1986; Ornelles et al., 1986; LubyPhelps et al., 1988; Isaacs and Fulton, 1987; Clegg et al., 1990; Laursen et al., 1990; Nickerson and Penman, 1991; Carter et al., 1993).

The cytoplasmic network, or cytomatrix, as viewed by whole-mount TEM of extracted cells, is formed from Factin filaments, microtubules, intermediate filaments and uncharacterized 2 to 3 nm filaments (Schliwa and van Blerkom, 1981). In the intact cell, the cytomatrix is locally differentiated into subcompartments that are architecturally and, by inference, functionally distinct. Stress fibers are one well-known example. In addition, exclusion of glycogen, ribosomes and organelles from the region surrounding the Golgi apparatus was observed by Mollenhauer and Morre (1978), who noted a corresponding ‘condensed’ appearance of the cytoplasm. Temmink and Spiele (1980) described four ultrastructurally different types of cytoplasm in nonmotile Swiss 3T3 cells, including a peripheral ‘cytocortex’ that contained a dense network of short filaments and was devoid of organelles, and a central ‘endoplasm’ that contains most of the microtubules, mitochondria and endoplasmic reticulum. Bridgman et al. (1986) distinguished ‘peripheral’ and ‘central’ cytoplasm with features similar to the cytocortex and endoplasm, respectively. Immunofluorescent staining revealed that the peripheral cytoplasm contained a meshwork of F-actin, but few microtubules. Correlative video-enhanced DIC microscopy showed that Brownian motion of small endogenous particles in the peripheral cytoplasm was highly restricted as if the particles were caged in a filament network. These peripheral cytoplasmic compartments resemble the leading lamellipodia of migrating cells and the axon growth cone, both of which lack large organelles and contain dense meshworks of actin filaments (Abercrombie et al., 1971; Small et al., 1982; Letourneau and Ressler, 1983).

Recently, we have reported that inert, fluorescent tracer particles with a hydrodynamic radius of 24 nm are excluded from a perinuclear compartment, and from compartments in the distal cytoplasm of adherent tissue culture cells, including the leading lamellipodia of migrating cells (LubyPhelps and Taylor, 1988). Because the tracer particles do not appear to interact with intracellular components, the most likely mechanism of exclusion is steric hindrance. Plausible steric mechanisms of exclusion include a physical barrier at the boundary between excluding and nonexcluding regions; a constriction in the height of the cell (either at the boundary or throughout excluding regions); molecular sieving by a gel-like cytomatrix; or close apposition of membranes, such as in the Golgi apparatus.

In this paper, we describe the architecture of size-excluding compartments as viewed by whole-mount TEM, immunofluorescence and specific fluorescent organelle staining. Comparison of our results with those of previous investigators leads us conclude that distal size-excluding compartments, the cytocortex described by Temmink and Spiele (1980), and the peripheral cytoplasmic compartments described by Bridgman et al. (1986) are one and the same. Non-excluding compartments correspond to the endoplasm or central cytoplasm. We also show that the Golgi apparatus is a size-excluding compartment within the larger perinuclear compartment, which is probably the centrosome. On the basis of the results of this study, we propose that the most likely mechanism for exclusion from the distal excluding compartments is molecular sieving by a filament network composed of F-actin and the actin-crosslinking protein, ABP280. Exclusion from the Golgi apparatus may be due to close apposition of cisternae in the stack, while exclusion from the centrosome is more likely to be due to a cytoskeletal structure, perhaps a basket of vimentin filaments.

All cell lines were obtained from the American Type Culture Collection (Rockville, MD). Media and tissue culture chemicals were from Gibco BRL (Grand Island, NY). Swiss 3T3 cells were grown in Dulbecco’s modified Eagle’s medium, 10% donor calf serum and 1% penicillin-streptomycin. CV1 cells were grown in Dulbecco’s modified Eagle’s medium, 10% fetal calf serum and 1% penicillin-streptomycin. PtK₁ cells were grown in Minimum essential medium, 10% fetal calf serum, 1 mg/ml sodium pyruvate and 1% penicillin-streptomycin. Cell cultures were maintained at 37°C in a humidified 5% CO₂ atmosphere.

F400 (Pharmacia Fine Chemicals, Piscataway, NJ) was activated and labeled with either tetramethylrhodamine isothiocyanate (Molecular Probes, Eugene, OR) or CY-5.18 (gift from Dr Alan Waggoner, Carnegie Mellon University, Pittsburgh, PA), as previously described (Luby-Phelps, 1988). The labeled polymers were fractionated by size exclusion chromatography on a 4.6 cm ×950 cm column of Sepharose CL-6B (Pharmacia Fine Chemicals) in 10 mM Tris, 50 mM KCl, 0.02% NaN₃, pH 8.0, at 20°C. Fractions of 5 ml were collected and dialyzed in water, lyophilized and stored at −20°C until used. Lyophilized Ficoll was reconstituted in injection buffer (2.5 mM PIPES, 0.05 mM MgCl₂, 0.05 mM KCl, pH 7.0). The mean hydrodynamic radius of particles in each size fraction was measured by fluorescence recovery after photobleaching as previously described (Luby-Phelps et al., 1987). A rhodamine-Ficoll fraction with a mean radius of 3 nm (Rh-F3) and a Cy5-Ficoll fraction with a mean radius of 24 nm (Cy5-F24) were used for this study.

Microneedles were prepared from Kwik-Fil Borosilicate Glass Capillaries (World Precision Instruments, Inc., Sarasota, FL) on a model 720 vertical pipette puller (David Kopf Instruments, Tujunga, CA). Microinjection was carried out using an Eppendorf Microinjector 5242 (Carl Zeiss Instruments, Thornwood, NY) with an MO-202 hydraulic micromanipulator (Narishige Co., LTD., Tokyo, Japan) on an Axiovert 35 microscope equipped with a 32 × LD Achrostigmat Ph1 objective (Carl Zeiss Instruments). Cells were maintained at ∼37°C during the injections using a KT model 5000 stage controller (Micro Devices, Inc., Jenkintown, PA). After injection, cells were rinsed 3 × in complete DMEM without phenol red and allowed to recover for 1-3 hours prior to imaging.

Coverslips 40 mm in diameter (Fisher Scientific, Custom Order cat no. 40Cir no.1, Pittsburgh, PA) containing injected cells were assembled into a 60 mm diameter modified Sykes-Moore chamber (Custom Scientific Inc., Dallas, TX) with a 0.5 mm thick sapphire window as the upper window of the chamber (Crystal Systems, Salem, MA). When used in conjunction with the KT model 5000 stage controller (Micro Devices), this chamber maintains the cultures at 37°C with a temperature differential between the edge and the center of the chamber of < 0.3 deg. C. Fluorescein fluorescence was viewed with a 485/20 nm bandpass excitation filter, 505 nm dichroic mirror and 542/45 nm bandpass emission filter. Rhodamine fluorescence was viewed with a 546/12 nm bandpass excitation filter, 565 nm dichroic mirror and a 590/40 nm bandpass emission filter. Cy5 fluorescence was viewed with a 646/19 nm bandpass excitation filter, 660 nm dichroic mirror and a 682/22 nm bandpass emission filter. DMB-ceramide staining was viewed with filter sets as described by Pagano et al. (1991). All filter sets were purchased from Omega Optical (Brattleboro, VT). Fluorescence was viewed using a 40 × Plan-Neofluar objective (NA 0.75) or a 100 × oil immersion Plan-Neofluar objective (NA 1.3) on an Axiovert 35 microscope (Carl Zeiss Instruments). A Photometrics Series 200 Cooled CCD camera (Photometrics LTD, Tucson, AZ) with input to a Perceptics 9200 image processor with μVax host (Perceptics, Knoxville, TN) was used to obtain digital fluorescence images. Exposure times were in the range of 3-10 seconds. Background images were obtained from areas of the coverslip containing uninjected cells. Following background subtraction, the numerator and denominator images were registered interactively, using the nucleus as a fiduciary marker, and the ratio was calculated using the floating point accelerator board of the Perceptics 9200. The image processor automatically rescales the images for display between 0 and 255 gray levels, while retaining the scaling factor and returning the actual ratio values when queried with the cursor or statistics functions.

For measurements of fluorescence intensity used to generate Fig. 9, rhodamine and Cy5 fluorescence images were transferred to a Macintosh IIfx computer (Apple Computer, Cupertino, CA). Regions of interest were outlined interactively and the grey-scale values at each pixel were tabulated using NIH Image v. 1.41. Regions of interest were chosen to be 3 to 4 pixels in height and 17 to 28 pixels in length, giving an area ≥ 68 pixels. Data tables were imported into CricketGraph v.1.3.2 (Cricket Software, Malvern, PA) for plots of Cy5 intensity vs rhodamine intensity.

Cells were plated for 18 hours prior to an experiment on 100 mesh gold finder grids (Ted Pella, Inc., Redding, CA) that were sandwiched between a 40 mm diameter coverslip and a sheet of carbon-coated Formvar (Electron Microscopy Sciences, Fort Washington, PA). A mixture of small Ficoll (Rh-F3) and large Ficoll (Cy5-F24) was microinjected as described above into cells that were growing on the Formvar in the open spaces of the grid. After recovery, cells were imaged as described above for no more than 20 minutes to minimize differences in morphology between the live and fixed cells. Following imaging, cells were quickly rinsed 3 × in filtered PHEM buffer (60 mM Pipes, 25 mM Hepes, 10 mM EGTA and 2 mM MgCl₂, pH 6.9) (Schliwa and van Blerkom, 1981) at 37°C and fixed in 2% glutaraldehyde (Ted Pella, Inc.) at 37°C for 15 minutes. The grid was then removed from the coverslip and rinsed 3 × in filtered PHEM at room temperature followed by 30 seconds in 1% aqueous OsO₄ (Ted Pella, Inc.). The brief osmication was sufficient to preserve membranes without creating artifactual actin filament networks (MaupinSzamier and Pollard, 1978). The grid was then rinsed 3 × in filtered distilled water, stained in 2% uranyl acetate for 10 minutes and rinsed again 3 × in filtered distilled water. The cells were dehydrated through an acetone series and critical point dried in a CPD 020 (Balzers Union, Hudson, NH) using bone-dry liquid CO₂ filtered through a water and an oil trap (Ris, 1985). Dry samples were carbon-coated and stored under vacuum in a desiccator until examined in the electron microscope. Stereo pairs for morphometry were taken at a nominal magnification of 20,000 with tilts of ± 20°off axis on a JEOL 1200 CX electron microscope.

Stereo pair negatives were digitized at 1024 × 1024 pixels using a Dage model 81 video camera (Dage MTI, Michigan City, IN) and an Androx ICS-400 imaging board (Androx Corp., Canton, MA) in a Sun 3/260 host (Sun Microsystems, Mountain View, CA). The images were displayed in stereo at a final magnification of ×138,000 on a Raster Technologies Model One/80 graphics terminal (1280 × 1024 pixel screen, two-monitor plus beam splitter stereo method, Alliant Computer Systems, Littleton, MA). Using STERECON (Marko et al., 1988) software, a stereoscopic cursor was used to find points in corresponding locations that were judged to be on the upper and lower surfaces of each excluding or non-excluding region in the image. These points were used to determine the thickness of the cell in each area. Five sets of points were averaged for each thickness measurement. The locations for measuring thickness were chosen, when possible, to avoid obvious ridges in the cell. To measure the mean diameter of void spaces in the cytomatrix, short 3-D line segments were drawn across gaps in the meshwork of fibers in each excluding or nonexcluding region. Twenty of these measurements were made for each area and averaged to obtain a mean pore size for the cytomatrix in each region.

Cells were cultured on 22 mm × 22 mm finder coverslips (Bellco Glass, Inc., Vineland, NJ) at least 18 hours prior to an experiment. Cells were injected with a mixture of Rh-F3 and Cy5-F24 and imaged as described above. Immediately after imaging, cells were quickly rinsed 3 × in PHEM at 37°C and fixed in 3.7% formaldehyde at 37°C for 15 minutes. Next, cells were rinsed 3 × in PHEM, for 5 minutes each at room temperature, permeabilized in 0.2% Triton X-100 for 90 seconds and then rinsed again 3 × in PHEM. The cells were exposed to the primary antibody (5-10 μg/ml) for 1 hour in a humidified chamber, rinsed 3 × in PHEM, for 5 minutes each at room temperature, then exposed to the secondary antibody (5 μg/ml) for 1 hour. In some cases, F-actin was labeled by placing 0.5 unit BODIPY phallacidin (Molecular Probes) on a coverslip along with the secondary antibody. Cells were rinsed 3 × in PHEM and were mounted in Fluoromount G (Fisher Scientific) with 2.5% DABCO to retard photobleaching. Anti-tubulin monoclonal antibody was a generous gift from Dr George Bloom, UT Southwestern Medical Center. Anti-vimentin (Sigma Chemical Co., cat. no. V5255) was a monoclonal antibody directed against human foreskin fibroblasts and reported to label vimentin in immunoblots and by immunofluorescence. Anti-αactinin (Sigma Chemical Co., cat. no. A5044) was a monoclonal antibody directed against bovine mammary gland epithelial cell cytoskeletons and shown to be specific for α-actinin by immunofluorescence and immunoblotting. mAb1 (a generous gift from Dr John Hartwig, Harvard University) was a mouse monoclonal antibody directed to the carboxy terminus of ABP280 (Gorlin et al., 1990). mAb1 does not recognize ABP280 well in mouse-derived cell lines, therefore all studies involving ABP280 were performed in CV1 cells and not Swiss 3T3 cells.

The procedure for staining the Golgi apparatus and endoplasmic reticulum of living cells was adapted from the procedure of Pagano et al. (1991). Cells were washed three times in serum-free medium, and then incubated for 5 minutes at 37°C in serum-free medium containing 2.5 μM DMB-ceramide (Molecular Probes, Inc.) and 0.34 mg/ml fatty acid-free BSA (Sigma). Following three more rinses in serum-free medium, the cells were incubated in serum-free medium containing 0.34 mg/ml fatty acid-free BSA for 45 minutes at 37°C before viewing by fluorescence microscopy.

The ultrastructure of excluding compartments was examined in both Swiss 3T3 and PtK₁ cells growing at sub-confluent densities. Cells grown on Formvar-coated gold finder grids were co-injected with Rh-F3 and Cy5-F24. Fluorescence ratio imaging was performed (see Materials and Methods) to obtain the pixel-by-pixel fluorescence intensity ratio of Cy5 to rhodamine in individual living cells. The intensity values in the resulting ratio image indicated the relative concentration of 24 and 3 nm particles in all regions of the injected cell, providing a map of regions that excluded 24 nm tracer particles, but not 3 nm tracer particles (Luby-Phelps and Taylor, 1988). These cells then were fixed immediately and processed for whole-mount TEM (see Materials and Methods). The injected cells were identified both by their position on the alphanumeric finder grid and by their morphology. Comparisons with uninjected cells on the same grid showed no apparent differences in ultrastructure and confirmed our previous findings that microinjected tracer particles are readily extracted from permeabilized cells even after fixation (Luby-Phelps et al., 1985, 1986, 1987; Luby-Phelps and Taylor, 1988). Membrane and protein structures such as the cristae of mitochondria, the endoplasmic reticulum, microfilaments, microtubules and coated vesicles were clearly defined in these whole-mount specimens (Fig. 1), suggesting that the ultrastructure of the cytomatrix was likewise well-preserved and informative of the cytoarchitecture of the living state.

The perinuclear excluding compartment was often recognizable as an electron-dense structure having the same projected outline as in the ratio image, but the specimens were too thick to observe the architecture of this region in any detail even in stereo micrographs. The distal excluding compartments were readily recognizable as regions devoid of membrane-bound organelles and having the same outline as in the fluorescence ratio image (Figs 2 and 3). Swiss 3T3 fibroblasts generally had a number of irregularly shaped excluding regions in the periphery (Fig. 2), while PtK₁ cells exhibited a broad circumferential band of exclusion at the periphery (Fig. 3). The latter morphology was also seen in CV1 cells by fluorescence ratio imaging, but this cell type was not included in the EM study. In both 3T3 and PtK₁ cells, the transition between excluding and non-excluding regions was quite abrupt, as it is in the ratio images of living cells. No physical barrier was observed at the boundary between excluding and non-excluding compartments, but the excluding compartments were usually thinner than adjacent non-excluding regions. The difference in height most likely reflects the natural contours of the cell rather than an artifact of collapse or uneven shrinkage during preparation for TEM, since the lower fluorescence intensity of Rh-F3 in these regions of living cells also indicated a reduced pathlength.

Distal excluding regions lacked microtubules, mitochondria, endoplasmic reticulum and other membrane-bounded organelles, except for an occasional coated vesicle. These organelles were found only in non-excluding compartments near the cell center, in channels traversing the distal excluding compartments, or grouped together in islands completely surrounded by an excluding compartment. The most prominent ultrastructural feature of the distal excluding compartments was a cytomatrix of numerous, long bundles of thin filaments running both radially and circumferentially, with a meshwork of more randomly oriented 5 to 10 nm filaments filling the space between them (Figs 2 and 3). On average, the void spaces in the cytomatrix of excluding compartments appeared significantly smaller than in adjacent non-excluding compartments.

To quantify the differences in the ultrastructure of the cytomatrix in excluding and non-excluding domains of the distal cytoplasm, the height of the cell and the size of void spaces in the cytomatrix were measured by stereo morphometry as described in Materials and Methods. Based on 20 measurements on each of seven stereo pairs, the average size of void spaces in excluding regions was 31 ± 9 nm s.e.m. In the adjacent non-excluding regions, the mean size of void spaces was 53 ± 13 nm s.e.m. The mean height of the cell in distal excluding compartments was 55 ± 26 nm s.e.m., while the mean height in adjacent non-excluding compartments was 128 ± 57 nm SEM. Mean values of the morphometric data from the two cell types were not significantly different.

We used the location of the Golgi apparatus as a marker to see whether the perinuclear size-excluding compartment corresponded to the centrosomal region of the cell. SW3T3 and CV1 cells were co-injected with Rh-F3 and Cy5-F24 as above and then stained with DMB-ceramide as described in Materials and Methods. DMB-ceramide is a fluorescent Ceramide analog that is taken up by the cells, transported to the endoplasmic reticulum, where it fluoresces green, and is subsequently concentrated in the transGolgi apparatus, where it fluoresces red (Pagano et al., 1991). Fluorescence ratio imaging using the ×100 objective showed that the perinuclear size-excluding compartment consisted of two domains: a compact, strongly excluding region closely apposed to the nucleus, surrounded by a larger, more diffuse region where the exclusion is less pronounced (Fig.4A). The compact, strongly excluding region coincided with the trans-Golgi as visualized by DMB-ceramide staining (Fig.4B). The surrounding region co-localized with a perinuclear accumulation of endoplasmic reticulum (Fig. 4C,D). It is unlikely that the exclusion from the Golgi apparatus was an artifact due to absorption of the Cy5 excitation light by the high concentration of DMB-ceramide in this organelle, since DMBceramide has no absorption band near 650 nm even at high concentrations of the probe in artificial membranes (Iain Johnson, Molecular Probes, personal communication). In addition, exclusion from the Golgi apparatus was evident in cells not stained with DMB-ceramide. Immunofluorescence revealed that the perinuclear excluding compartment also coincided with intense staining for the intermediate filament protein, vimentin (Fig.5).

To determine whether the filamentous cytomatrix observed in distal excluding regions was composed of known cytoskeletal structures, we localized tubulin, vimentin, Factin and two actin-binding proteins, ABP280 and nonmuscle α-actinin by immunofluorescence after excluding compartments had been mapped by fluorescence ratio imaging as above. In both Swiss 3T3 and CV1 cells, tubulin was absent or scarce within excluding regions, although microtubules frequently skirted the boundaries between excluding and non-excluding compartments or ran down the center of a non-excluding channel (Figs 6A,B). Vimentin filaments also were rarely observed within distal excluding regions (Fig. 5). In contrast, F-actin (Fig. 7), ABP280 (Figs 7 and 8) and α-actinin (Fig. 8) were present throughout the cytoplasm, including distal excluding compartments. F-actin and ABP280 co-localized on stress fibers and smaller filament bundles. Diffuse fluorescent staining for these antigens was observed also in the cytoplasm between filament bundles (Fig. 8). In contrast, α-actinin staining was restricted to punctate decoration of F-actin containing filament bundles, with a periodicity of ≈0.4 μm (Fig. 8). This spacing was not different in excluding compartments compared with non-excluding compartments. In general, the diameter of the filament bundles in excluding compartments was less than or equal to the diameter of filament bundles in non-excluding compartments.

The reduced cell height (and therefore volume) in distal excluding compartments compared with non-excluding compartments suggested that F-actin and ABP280 might be more concentrated in these regions of the cytoplasm (Taylor and Wang, 1980). We confirmed this by using RhF3 as a volume indicator to normalize the immunofluorescence images (Fig. 7): images of the immunofluorescence localization of F-actin or ABP280 were divided by the image of the distribution of Rh-F3 in the living cell before fixation. Only cells whose gross morphology was unchanged after fixation were used.

To see whether the reduced height of the cell in distal excluding compartments was the primary mechanism for exclusion of larger tracer, we plotted the fluorescence intensity from the large tracer particles (Cy5) against the fluorescence intensity from the small tracer particles (rhodamine) for individual pixels in the distal excluding and non-excluding compartments of three cells on the same coverslip. Regions of interest were chosen to span several micrometers of the cell within each type of compartment in order to sample the widest possible range of pathlengths. In non-excluding compartments, Cy5 intensity scaled linearly with rhodamine intensity (Fig. 9A), indicating that the fluorescence intensity from both sizes of particle was simply a function of the pathlength (cell height). In contrast, Cy5 fluorescence did not scale with rhodamine fluorescence in distal excluding compartments (Fig. 9B). Importantly, in one of the three cells, Cy5 fluorescence is several-fold lower in excluding compartments at rhodamine intensities overlapping those of the thinner regions of nonexcluding compartments (Fig. 9B, squares).

We have examined the cytoarchitecture of subcellular compartments that exclude large inert tracer particles in order to gain insight into the possible mechanism of exclusion and the cellular structures that are responsible. The perinuclear size-excluding compartment clearly is the centrosome, as indicated by the presence of the Golgi apparatus within this compartment and also by its characteristic location between the nucleus and the leading edge of polarized, migrating cells (Luby-Phelps et al., 1988). A structure that is visible by EM and that stains intensely for vimentin is also present in this region, and we previously have shown that clusters of mitochondria radiate out from this position in migrating cells (Luby-Phelps et al., 1988). Without optical sectioning of double-stainined material, we have insufficient information regarding the relative arrangement of these organelles and the vimentin cytoskeleton in the centrosome to do more than speculate regarding the mechanism of exclusion from this region. However, two possible mechanisms are readily suggested by the present evidence. Pagano has shown that the discrete tubular structures seen by DMB-ceramide staining often represent small stacks of trans-Golgi cisternae (Pagano et al., 1989, 1991). Inspection of the electron micrographs in these references yields an estimated spacing of 12 to 23 nm between the cisternae in each stack: close enough to exclude Cy5-F24 but not RhF3. This mechanism is unlikely in the case of the surrounding domain that coincides with the endoplasmic reticulum because the membranes in this region are less tightly packed. The observation of intense vimentin staining in this region suggests that large tracer might be excluded by a network or basket of intermediate filaments. It is possible that exclusion from both subdomains of the perinuclear compartment may reflect some local differentiation of the cytomatrix that integrates the centrosomal region structurally and functionally.

On the basis of their ultrastructure, we conclude that distal size-excluding compartments are synonymous with the ‘cytocortex’ previously described by Temmink and Spiele (1980) and the ‘peripheral’ cytoplasm described by Bridgman et al. (1986). We observed no physical barrier at the boundary between distal excluding and non-excluding compartments that would account for exclusion of large tracer particles. Distal excluding compartments were generally thinner than adjacent non-excluding domains, and it is possible that constriction may in some cases account for exclusion. However, two lines of evidence suggest that this is not the primary mechanism of exclusion. Allowing for shrinkage of the cell during preparation for whole-mount TEM, 55 nm can be taken as the lower limit for the mean height of distal excluding compartments. This height is just sufficient to permit entry of Cy5-F24. In addition, the intensity of Cy5 fluorescence from the large particles does not scale linearly with the intensity of rhodamine fluorescence from the small particles in distal excluding compartments, as it does in adjacent non-excluding compartments (Fig. 9). This suggests that there is no relationship between pathlength and exclusion from the distal excluding compartments.

The cytoarchitecture of distal excluding compartments is consistent with a third hypothesis that exclusion is due to inability of large tracer particles to percolate through the filamentous cytomatrix. In whole-mount TEM specimens, excluding compartments are filled with a dense cytomatrix comprising numerous long filament bundles interconnected by a meshwork of more randomly oriented filaments. The mean size of void spaces in the meshwork of filaments between the fiber bundles in the whole-mount TEM specimens is only about half the size of void spaces in the cytomatrix of adjacent non-excluding regions. The 5 to 10 nm diameter of the filaments in the bundles suggests they are F-actin (Engelman and Padron, 1984). This conclusion is supported by the immunofluorescence localization of Factin, ABP280 and α-actinin on filament bundles of similar size, number and orientation in excluding compartments. It is likely that the meshwork of filaments between filament bundles contains F-actin and ABP280, since diffuse staining for these antigens was observed between filament bundles in immunofluorescent specimens (see also Bridgman et al., 1986). In vitro, ABP280 crosslinks F-actin into isotropic networks (Niederman et al., 1983). A role for an F-actin/ABP280 network in the exclusion of large particles from distal cytoplasmic compartments is supported by evidence in the literature. These compartments ultrastructurally resemble the leading lamellipodium of a migrating cell, which contains a dense meshwork of F-actin filaments (Small et al., 1982). In the mammalian macrophage, ABP280 has been localized to crosslinks or overlaps between F-actin filaments in the meshwork (Hartwig and Shevlin, 1986). Recently, it was reported that human melanoma cell lines deficient or lacking ABP280 do not make lamellipodia that exclude organelles (Cunningham et al., 1992). This phenotype was restored by transfecting the ABP280⁻ cell lines with a mammalian expression vector containing the gene for ABP280 (Cunningham et al., 1992). The distinct ultrastructure of distal excluding and nonexcluding compartments raises the intriguing possibility that they represent functionally specialized subdomains of the cytoplasm. The restriction of microtubules and membrane-bounded organelles such as mitochondria and the endoplasmic reticulum to non-excluding compartments suggests that not only vesicle traffic (Bridgman et al., 1986) but also protein synthesis may occur exclusively in these domains, which are often narrow channels radiating from the cell center. Division of the cytoplasm of cultured embryonic chick sensory neurons into ‘translational’ and ‘expressional’ cytoplasm has previously been proposed, based on the absence of ribosomes in the distal portions of neurites (Baas et al., 1987). Two recent observations support this idea. Staining of RNA with fluorescent vital dyes shows that in living cultured cells ribosomes are localized in channels similar in morphology to non-excluding compartments (Terasaki, 1991). Occasionally transport of the mRNA for β-actin from the nucleus to the distal cytoplasm appears to occur in narrow pathways similar in appearance to nonexcluding channels (Kislauskis and Singer, 1992; Rob Singer, personal communication). The dense cytomatrix and the lack of microtubules in excluding compartments suggest that endocytosis also may occur preferentially over the non-excluding domains, although we and others have observed occasional coated vesicles in excluding compartments (Mollenhauer and Morre, 1978). Restriction of these cell functions to a cytoplasmic subcompartment might increase their efficiency by reducing the volume accessible to diffusible reactants. Additionally, channeling of transport might allow rapid delivery of organelles to specific sites in response to external signals received at the plasma membrane. For example, a recent abstract reports that when two growth cones meet in culture, organelle and microtubulecontaining channels invade both growth cones and extend to the site of contact (Lin and Forscher, 1991). Restriction of mitochondria and endoplasmic reticulum to subdomains of the cytoplasm might also lead to microcompartmentation of ATP and Ca²⁺. Independent evidence for microcompartmentation of these substances, as well as other small solutes, such as glycolytic intermediates, fatty acids, amino acids and nucleic acid precursors, can be found in the literature (for general reference, see Jones, 1988). Microcompartmentation of ATP and Ca²⁺ could be a mechanism for spatial control of various metabolic processes, while changes in microcompartmentation might be a feature of temporal regulation.

The ultrastructure of distal excluding compartments and the rheological properties of actin-ABP280 networks in vitro suggest that excluding compartments might be specialized for mechanical rigidity. Actin-ABP280 gels are similar to covalent networks in that they are relatively rigid and elastic, exhibiting no creep below a critical strain at which the actin filaments break (Janmey et al., 1990; Zaner, 1986). The slow turnover of crosslinks implied by these properties would account for the sharp percolation cutoff and the static character of excluding compartments, which persist unchanged over the course of several hours in wellspread cells. The only available measurement of the mechanical properties of an excluding compartment in vivo supports the idea that these compartments have a mechanical function. Felder and Elson (1990) have recently demonstrated that the leading lamellipodia in migrating fibroblasts are stiff and elastic even after detachment from the substratum as ruffles. Our preliminary observations suggest that excluding compartments in non-motile well-spread cells are the remnants of lamellipodia that were extended during spreading or locomotion and subsequently invaded by non-excluding channels. The mechanical advantages of extending a very thin, rigid lamellipodium through a relatively viscous medium are intuitively obvious. In the spread cell, residual excluding compartments might be responsible for maintaining the flattened morphology of the cell, perhaps coinciding with sites of close contact with the substratum. In that case, remodelling of the cytoarchitecture of excluding compartments might accompany physiological changes in cell shape. Among the biologically important examples of regulated changes in cell shape are the polarization of cells at the edge of an experimental wound, the extension and consolidation of axon growth cones, the retraction of endothelial cells in response to inflammatory agonists, and the response of serum-deprived cells to growth factors.

We are grateful to Fred Lanni and Lans Taylor for helpful discussions, to William Donzell and Dennis Bellotto for assistance and guidance in electron microscopy, and to John Hartwig for the gift of ABP280 antibody, as well as helpful discussions of the data. We also thank Malú Tansey, Helen Yin and Jim Stull for reading this manuscript, and Iain Johnson and Josh Stahl of Molecular Probes, Inc. for providing information about the absorption spectrum of DMB-ceramide. We are also indebted to the anonymous reviewer of this manuscript for suggesting the measurements detailed in Fig. 9. This research was supported by NSF DCB-89 16421 to Dr Luby-Phelps. Dr Luby-Phelps is an Established Investigator of the American Heart Association. The NIH Biological Microscopy and Image Reconstruction Resource, at the Wadsworth Center in Albany, NY, is supported by Biotechnological Resource grant RR01219.