Axolotl pronephric duct cell migration is sensitive to phosphatidylinositol-specific phospholipase C

During the course of development, cells undergo directed migrations that are required for the generation of the embryo’s three-dimensional form. What directs a cell to its appropriate destination, which is often quite distant from its site of origin? We have been investigating the migration of pronephric duct cells in embryos of the axolotl (Ambystoma mexicanum) at both the cellular and molecular levels (Zackson & Steinberg, 1986, 1987, 1988; Gillespie et al. 1985; Steinberg & Poole, 1982; Poole & Steinberg, 1981, 1982). Pronephric duct cell migration is a favorable system in which to investigate this question because the system consists of a population of cells of a single type that migrate along a defined pathway towards a single target and is accessible to a variety of experimental approaches.

The pronephric duct primordium first appears as a nest of cells ventral to the anterior somites shortly after the completion of neurulation. As development proceeds, the primordium elongates along the ventral edge of the newly formed somites. This elongation involves a rearrangement of pronephric duct cells, and does not involve either cell division or the elongation of individual cells. Migration and rearrangement of the pronephric duct cells continues until cells at the posterior tip reach the target organ, the cloaca. Although motile, pronephric duct cells remain adherent to one another during the course of the migration. In normal development, pronephric duct elongation is in synchrony with the craniocaudal wave of somitogenesis: the posterior tip of the duct is normally two somite widths anterior to the last-formed somite fissure.

A series of grafting experiments have shown that the information guiding the pronephric duct cells is local to their migration substratum and displays many formal properties expected of a gradient of adhesiveness (Poole & Steinberg, 1982; Zackson & Steinberg, 1986, 1987). It has also been shown that the guidance information is not restricted to the pronephric duct pathway; the neighbouring lateral mesoderm can also both support and direct pronephric duct cell migration. Pro-nephric duct cells grafted to the lateral mesoderm migrate dorsocaudad towards the pronephric duct pathway proper, then turn sharply caudad and migrate towards the cloaca upon reaching the pathway. In addition, we have shown that the guidance information is not restricted to pronephric duct cells: cranial neural crest cells, which normally migrate in a different region of the embryo, will exhibit directed migration when grafted ectopically to the lateral mesoderm or pronephric duct pathway (Zackson & Steinberg, 1986). Because cranial neural crest cells are highly motile and migrate as individual cells (rather than as a coherent population), we have used grafted cranial neural crest cells as probes for the distribution of the guidance information. Two important features of this system revealed by grafts of these cells are that cranial neural crest cells migrate unidirectionally upon the pronephric duct pathway but bidirectionally upon the pronephric duct primordium itself. From the trajectories of grafted pronephric duct and cranial neural crest cells we have derived a map of the distribution of the guidance information in the area examined.

We have recently found (Zackson & Steinberg, 1988) that the distribution of the cell surface enzyme alkaline phosphatase corresponds closely to this map: (1) it is distributed as a gradient on the lateral mesoderm increasing towards the pronephric duct pathway; (2) it is reduced or absent on the presomite and early somite mesoderm as well as on the overlying epidermis; (3) it is on the pronephric duct itself; and (4) in some embryos a gradient of this molecule on the pronephric duct pathway proper can be observed. Although it has long been known that alkaline phosphatase has a patterned distribution in embryos (Moog, 1944; Brachet, 1946), little is known about its physiological function or possible roles in embryogenesis.

In initial experiments not presented here, we applied to embryos the alkaline phosphatase inhibitor levamisole to determine whether it would inhibit pronephric duct migration. Levamisole was indeed found to do so, and in a concentration range similar to that required for it to inhibit the endogenous alkaline phosphatase activity. However, levamisole may interact with other proteins in addition to alkaline phosphatase (e.g. cholinergic receptors of nematode muscle: Harrow & Gration, 1985; Lewis et al. 1987a,b) and we found that at concentrations slightly higher than that required for minimal inhibition of pronephric duct cell migration, other abnormalities became apparent. For example, at such concentrations the epidermis, which displays little alkaline phosphatase activity, dissociates into single cells. Hence, because of the possibility of side effects of the drug, we regarded these observations as merely suggestive.

Recent investigations have shown that alkaline phosphatase is anchored to cell surfaces not by a transmembrane peptide domain but rather via a covalent linkage to a phosphatidylinositol glycan (PI-G; Howard et al. 1987; Ikezawa et al. 1976; Low & Zilversmit, 1980; Low & Saltiel, 1988). Because the PLG linkage is the only means by which alkaline phosphatase is held to the cell membrane, alkaline phosphatase (and other similarly anchored proteins) can be released from cell surfaces by the enzyme phosphatidylinositol-specific phospholipase C (PIPLC).

In this paper, we test the effects of PIPLC on pronephric duct cell migration. In order to conduct these experiments, we have developed a novel slow-release bead system for introducing macromolecules into the embryo in a focal sustained manner. PIPLC is shown both to inhibit pronephric duct cell migration and to remove alkaline phosphatase from cell surfaces. Our results support the hypothesis that at least one PL G anchored molecule, possibly alkaline phosphatase, is involved in the directed migration of pronephric duct cells.

15 % polyacrylamide was prepared from a standard acrylamide stock solution (ratio of acrylamide to bis 30 %:0·8 %). The materials for the gel were mixed in the following volumes, in order: distilled water, 4-8 ml; acrylamide stock solution, 5 ml; TEMED, 100 μ1; 10% ammonium persulphate, 100 μl. After polymerization, the gel was coarsely chopped with a weighing spatula and allowed to remain overnight in distilled water. It was then ground by passing it through a series of nylon monofilament meshes of decreasing aperture size: a series consisting of aperture sizes 590μm, 420μm, 350 μm, 250 μm, 149 μm, 105 μm, and 53μm was used. Each mesh was cut to fit a Millipore filter holder and gels were forced through with the aid of a glass syringe. The resulting beads were again soaked in distilled water overnight. The beads were then spun at 720g for 30s, and the supernatant, which contained bits of polyacrylamide too small to be used, was discarded.

The procedure for base-catalysed hydrolysis is modified from that of Tanaka (1981). Beads were placed in a beaker with a large excess of 10% ethanolamine. The solution was then heated to 54 °C, and held at that temperature with stirring (on a hot plate kept under a fume hood) for 19 h.

Because the beads are transparent, it is necessary to dye them so that they can be seen both during operations and upon examination of results. The Hypa beads were washed by two cycles of suspension in water, followed by suspension in acetone. Acetone-dried beads were rehydrated in 40miu-NaCl, 50 mm-borate pH 9-3. Isopropyl alcohol was then added to 10%.

Tetramethylrhodamineisothiocyanate (TRITC; Research Organics, Inc.), dissolved at 40_μg₍μl⁻¹ in DMSO was added at 100 μ1 dye solution per ml beads. After 12 h, the beads were pelleted and washed with distilled water followed by acetone. After resuspension in 10 ml distilled water, 10 mg of the dye Neutral Red was added. After Ih, the beads were pelleted and washed through three cycles of acetone and water suspension. After three days in water, (during which time Neutral Red continued to diffuse from the beads), the beads were suspended in acetone, the acetone was removed and the beads were allowed to air dry. The neutral red permits visualization of the beads under the dissecting microscope, while TRITC staining permits visualization by epifluorescence illumination of sectioned specimens in the compound microscope.

Solutions of proteins to be tested were prepared as described below. A few drops of solution were placed in a watch glass held in a moist chamber and dried beads were dropped into the solution. Tungsten needles were used to agitate the beads when necessary; the beads were allowed to remain in protein solution for at least 30 min prior to use.

Horseradish peroxidase (HRP, type VI salt-free powder, Sigma) was dissolved in 15% Hepes buffered Steinberg solution (HBSt; Zackson & Steinberg, 1986) at high concentration (approx. 20 mg ml⁻¹). Trypsin (type IX, Sigma) was dissolved in HBSt at I mg ml⁻¹. Phosphatidylinositol-specific phospholipase C (PIPLC), a generous gift of Dr Martin Low, Columbia University, was previously assayed in his laboratory as having an activity of 1000 μmol min⁻¹ ml⁻¹, using [³H]phosphatidylinositol as substrate, and was judged to be over 90% pure by SDS-PAGE. This enzyme was supplied in 50% glycerol and 0·02% sodium azide, which were removed before use by the following method: 50 μ1 of enzyme was added to 2 ml of 15 % HBSt. This solution was concentrated by spinning in a Centricon 10 tube (10×10³M_r cutoff, Amicon Corp.) at 4300g until most of the solvent had passed through, leaving about 50 μl of retentate. 15 % HBSt was added to the retentate to a volume of 2 ml, and the centrifugation was repeated. 10 mg of BSA (Sigma, fraction V) was added to 2 ml of 15 % HBSt. The dissolved protein was concentrated in a Centricon 10 tube, diluted, and concentrated a second time.

A sample of approx. 180 Hypa beads loaded with HRP was quickly pelleted and resuspended twice with 15 % HBSt. After the final pelleting, the beads were resuspended in 0·2 ml of 15 % HBSt in a small column fitted with a porous glass frit. In order to reduce turbulence and flow around the beads, glass wool was pushed into the column and positioned so that the flow of liquid was broken up by the glass wool. Approximately 5 ml of 15 % HBSt was quickly pumped through the column with the aid of a peristaltic pump; this solution was discarded. 15 % HBSt was then pumped through at a rate of approx. 0·9 ml h⁻¹ for 25 h; fractions were collected over one hour intervals. As a control, a solution of HRP was applied to the column with no Hypa beads present. HRP activity was measured in the collected fractions by distributing 100 μl samples (or diluted aliquots of initial fractions) to individual wells of a 96-well plate. To each well was then added 100 μl HRP substrate solution: I mg ml⁻¹ o-phenylene diamine; 0·75μl ml⁻¹ 30% H2O2; 100mm-citric acid/sodium citrate, pH 4·5. The absorbance at 414 nm of each well was read at two minute time intervals on a BioRad ELISA plate reader. (The ELISA plate reader was blanked against air). The change in absorbance over time was plotted to determine enzyme activity. A standard curve was generated from a dilution series of a stock solution of known HRP concentration. From this curve the amount of active HRP in each sample could be calculated. The data points presented in Fig. 1 represent amounts of HRP (calculated from the standard curve) present in 100 μ1 samples removed from 0·9 ml fractions.

Embryos were cultured and operations conducted utilizing the methods described previously (Zackson & Steinberg, 1986, 1988). Except where noted, all operations were conducted upon embryos of approx. Schreckenberg & Jacobson (1975) stage 24 (approximate stage of onset of pronephric duct cell migration) and the operated embryos were allowed to develop until unoperated siblings reached approx, stage 30 (when pronephric duct cell migration is complete). Unless otherwise noted, beads were implanted by the following method: with the aid of an electrolytically sharpened tungsten needle, a small slit was made in the epidermis either over the lateral mesoderm or else over the posterior presomite mesoderm. A bead less than 100 μm in its longest dimension (approximately the size of a somite or smaller) was selected for implantation. The bead was carried to the slit with forceps, then pushed between the epidermis and mesoderm using a blunt-ended tungsten needle (while holding up the epidermis with forceps) until it was positioned on the presomitic mesoderm, dorsal and caudal to the posterior tip of the pronephric duct primordium. A glass bridge was then placed over the slit with gentle pressure for 40–60min.

Upon reaching the appropriate stage of development, the embryos were fixed and their epidermis was removed locally. They were then developed for alkaline phosphatase activity utilizing BCIP as a substrate as described previously (Zackson & Steinberg, 1988). Because embryos harbouring PIPLC-loaded beads could be distinguished easily from controls harbouring BSA-loaded beads (the former were considerably shorter), experimental and control embryos were all fixed, rinsed and developed for alkaline phosphatase activity in the same vessel (except where noted) to minimize possible artificial differences in alkaline phosphatase staining intensities.

Embryos were prepared for scanning electron microscopy as previously described (Zackson & Steinberg, 1986) except that the fixative buffer used was the same as that for preserving alkaline phosphatase activity (Zackson & Steinberg, 1988).

We describe here a method for releasing macromolecular reagents (such as enzymes and antibodies) into amphibian embryos in a focal sustained manner utilizing a novel carrier material, hydrolysed polyacrylamide (‘Hypa’). The detailed procedure for preparation of these beads is presented in Materials and Methods. When acetone-shrunken Hypa beads are reswollen in the presence of macromolecules, these are taken up by the beads. When a bead is transplanted to a new environment (such as an embryo), the loaded macromolecules diffuse from the bead.

The efficacy of Hypa beads in vitro was studied utilizing the easily assayed enzyme horseradish peroxidase (HRP). Fig. 1 illustrates the kinetics of HRP release into 15 % HBSt over a 25 h period from a short column containing Hypa beads. As a control, a dilute solution of HRP was applied to the column without any beads present and liquid pumped through as before These data indicate that HRP continues to be released from the beads for at least 25 h. In contrast, HRP applied to the column in the absence of any Hypa beads is quickly diluted away; the amount of HRP coming off the column drops by about two orders of magnitude within a 4h time span. Hence, release of the loaded enzyme in vitro is sustained over a time span required for continuous application of a reagent during the stages of pronephric duct cell migration in vivo.

The efficacy of Hypa beads in vivo was studied in a model experiment utilizing the proteolytic enzyme trypsin. Beads loaded with trypsin were implanted between the epidermis and mesoderm of stage-24 albino axolotl embryos. Trypsin released from an implanted bead caused visible damage in the vicinity of the bead within minutes of the operation. Damage was still readily apparent after 24 h.

When Hypa beads stained with RITC and neutral red and loaded with BSA or HRP were implanted into recipient embryos, development proceeded normally (15/17 cases). Of these, in 9/11 cases in which the bead was placed directly upon or immediately adjacent to the pronephric duct pathway, the pronephric duct cells migrated around the beads upon the lateral mesoderm. In two cases, a bead placed directly upon the pronephric duct pathway blocked pronephric duct cell migration. Fig. 2 provides an example of an embryo in which the bead partially obstructed the duct pathway. The pronephric duct cells migrated around the obstruction, returned to their pathway and continued migration towards their target. In addition, we have allowed some embryos containing Hypa beads to develop to swimming tadpole stages; they do so with no obvious developmental defects. Hence, implanted Hypa beads by themselves appear to be non-toxic to axolotl embryos and do not interfere with normal development.

The enzyme phosphatidylinositol-specific phospholipase C (PIPLC) specifically releases PI-G anchored proteins from cell surfaces (Low, 1987). In order to determine whether a PI-G anchored protein is involved in pronephric duct cell migration, individual Hypa beads loaded with PIPLC were implanted upon the presomitic mesoderm close to the pronephric duct pathway in stage-24 axolotl embryos. In two separate experiments (utilizing different spawnings of embryos and different preparations of beads), the embryos were permitted to develop until sibling controls reached approx, stage 30 or stage 36. In 26/26 cases, the pronephric duct cells ceased migration at a point anterior to the bead. 11 of these embryos were also peeled and examined on the unoperated side; in 10 cases, the pronephric duct cells ceased migration on both sides of the embryo.

Fig. 3 presents both light and scanning electron microscope (SEM) views of an embryo which had received a PIPLC-loaded Hypa bead. This embryo is a sibling of the control embryo displayed in Fig. 2. The embryos were fixed and peeled under identical conditions, and developed for alkaline phosphatase activity simultaneously. A comparison of these two embryos reveals several important features. First, PIPLC completely inhibits pronephric duct cell migration; the tip of the duct is three somite widths anterior to the bead in the PIPLC-treated embryo, little advanced beyond its position at the time of the operation. Second, a few ‘loose’ cells of unknown origin are present along the pronephric duct pathway in the vicinity of the PIPLC bead. Third, when stained for alkaline phosphatase activity, almost no blue colour develops in the PIPLC-treated embryo after 30 min, a time sufficient for blue colour to develop in the control. Fourth, the PIPLC-treated embryo has failed to straighten out and elongate. Fifth, somite fissures continue to form in both the PIPLC-treated and control embryos. Scanning electron microscopy permits accurate counting of somite fissures (Armstrong & Graveson, 1988). Of the embryos viewed in the SEM, three PIPLC-treated embryos displayed 16, 18 and 19 somite fissures, whereas two control sibling embryos containing BSA-loaded beads displayed 16 and 18 somite fissures (counting from the somite fissure dorsal to the anterior end of the pronephric duct), indicating that this morphogenetic cell rearrangement is not inhibited by the presence of PIPLC.

The PIPLC-loaded Hypa beads were remarkably effective at removing alkaline phosphatase activity. Fig. 4 shows a PIPLC-treated embryo which was fixed at approx, stage 30 and developed for alkaline phosphatase activity for 18 h. As indicated in the figure, regions far from the bead develop intense blue colour, whereas, in the vicinity of the bead, activity is still virtually undetectable. Indeed, a gradient of alkaline phospha-tase activity is apparent on the lateral mesoderm and pronephric duct, alkaline phosphatase activity increasing with distance from the bead. Even the normally heavily staining trunk-level neural tube shows little activity in the vicinity of the bead. Fig. 5 shows a pair of sibling embryos cultured to approx, stage 36 and developed for alkaline phosphatase activity for 16 h. The upper embryo received a control bead loaded with BSA; the lower embryo received a PIPLC-loaded bead. On both sides of the PIPLC embryo, not only is overall alkaline phosphatase activity markedly reduced, as indicated by the loss of blue colour development, but pronephric duct cell migration is inhibited as well.

In our preceding paper (Zackson & Steinberg, 1988), we proposed that in a morphogenetic system in which cells migrate in response to local environmental guidance cues (such as differences in adhesiveness), one or more ‘cell guidance associated molecules’ should be distributed in the embryo in a molecular prepattern corresponding to the distribution of the cell guidance information. Based upon migration trajectories of grafted cells, we were able to develop a map of the expected distribution of a molecule involved in the guidance of pronephric duct cells in the axolotl embryo. In addition, we identified a molecule that displayed a patterned distribution in the axolotl embryo closely corresponding to that expected for a molecule involved in pronephric duct cell guidance. Surprisingly, this molecule was found to be the phosphatidylinositolglycan anchored cell surface enzyme alkaline phosphatase. The present paper addresses the question of the possible functional involvement of such a PI-G anchored molecule in cell guidance.

Macromolecules of known activity and specificity, such as enzymes and antibodies, are potentially useful reagents for studying molecular mechanisms of development. However, the use of such macromolecular reagents requires a method for introducing them into the embryo. A slow-release system was therefore developed that permitted focal sustained release of macromolecular reagents at chosen sites in the embryo.

The major properties required of a slow-release vehicle are that it be (1) non-toxic to the embryo and (2) able to release the loaded reagent at a sufficiently high concentration over a time course corresponding to that of the developmental process under investigation. We have developed the Hypa bead system, which appears to meet both of these criteria in both in vitro and in vivo tests. HRP loaded into beads continues to diffuse out for 25 h in vitro, and control beads do not appear to interfere with normal development. Previous studies on the chemical and physical properties of polyacrylamide (Tanaka, 1978, 1981; Tanaka et al. 1980) have shown that when subjected to base-catalysed hydrolysis, many amide groups within the polyacrylamide matrix are converted to carboxylates. This chemical modification causes the polyacrylamide to swell to many times its initial volume, due to a complex set of physical interactions. In addition, acetone causes the Hypa to collapse to a small fraction of its original volume, this shrinkage being completely reversible upon rehydration. The volume ratio of hydrolysed polyacrylamide in water versus acetone is approximately 350 to 1 (Tanaka, 1981; Tanaka et al. 1980). Hence, beads dried in acetone undergo an enormous expansion upon rehydration and molecules dissolved in the rehydration solution are taken up by the beads. The sustained release of loaded proteins possibly involves two different properties of the beads. First, the Hypa might be acting simply as a sponge, due to its porosity and surface area. Second, the carboxylate groups generated during the alkaline hydrolysis of the beads are negatively charged, and therefore could provide cation exchange sites which would bind protein via electrostatic interactions. The salt present in the buffer or embryonic environment might slowly drive off the bound protein.

We expect that Hypa beads will prove useful for the introduction of many different reagents, of both low and high molecular weight, including antibodies, into embryos of many species.

Because the molecule whose distribution correlates with the pronephric duct guidance information was identified as the PI-G anchored cell surface enzyme alkaline phosphatase, and because preliminary experiments revealed that the alkaline phosphatase inhibitor levamisole blocked pronephric duct migration, we chose the enzyme phosphatidylinositol-specific phospholipase C (PIPLC) as a reagent for investigating the possible functional involvement of a PI-G anchored protein in cell guidance. PIPLC specifically releases from cell surfaces alkaline phosphatase and other proteins utilizing this linkage (Low, 1987; Howard et al. 1987). If a PI-G anchored protein is involved in cell guidance, then its removal from the cell surface should interfere with the directed migration of pronephric duct cells.

As noted above, pronephric duct cell migration is indeed inhibited by this enzyme, whereas somite fissure formation, a concurrent morphogenetic rearrangement of neighbouring cells displaying little if any alkaline phosphatase, proceeds apparently undisturbed in the presence of this enzyme. The enzyme also inhibits embryo elongation. Although poorly characterized, this latter process is likely to involve the rearrangement of cells of the lateral plate, which display alkaline phosphatase. Somite fissure formation thus serves as an internal control in these experiments: the failure of PIPLC to interfere with this process indicates that PIPLC is not simply interfering with morphogenesis or some related cell function in a general sense but rather is selectively affecting a subset of morphogenetic events. These observations also argue against the possibility that the observed effects of applied PIPLC are due to an unknown contaminant. For example, a nonspecific phospholipase would be expected to be toxic to all cells, whereas even in the vicinity of the implanted beads there is no evidence of unrestricted damage.

Staining of PIPLC-treated embryos for alkaline phosphatase activity reveals diminished activity at distances surprisingly far from the site of implantation. Diminished activity is observed over 1mm away from the bead, and on both operated and unoperated sides of the embryo. While this observation is most likely explained by diffusion of enzyme over this distance, we cannot rule out the possibility that the enzyme is acting directly only over a shorter distance, and that the locally affected cells transmit signals that cause a decrease in alkaline phosphatase expression in more distant cells. However, the minimal amount of colour development observed in the vicinity of the bead even after 18 h in the substrate solution attests to the potency of the PIPLC.

Although we demonstrate that little alkaline phosphatase activity remains on cells within a certain distance of an implanted bead containing PIPLC, this experiment does not prove that alkaline phosphatase is the mediator of pronephric duct cell guidance. It remains possible that one or more other PI-G anchored molecules are functioning in the guidance of pronephric duct elongation and that they are coexpressed with alkaline phosphatase; or that PIPLC’s inhibitory effects here are upon the process of migration itself rather than upon its guidance. These experiments do, however, provide strong evidence that at least one species of PI-G anchored molecule plays an active role in the directed migration of the pronephric duct.

In an earlier investigation of the axolotl pronephric duct guidance information, Gillespie et al. (1985) found that brief exposure of the mesoderm to trypsin in the absence of calcium ions not only blocked pronephric duct cell migration but also caused a local disruption of somitogenesis. However, exposure of the mesoderm to trypsin in the presence of calcium ions did not block pronephric duct cell migration but still caused a disruption of somitogenesis. Their observations suggest that there exists at least one trypsin-sensitive, calcium-protectable component of the pronephric duct guidance information. A comparison of the cellular responses to trypsin in the presence of calcium and to PIPLC demonstrates that while both treatments differentially affect pronephric duct and somite morphogenetic movements, they do so with opposite sensitivities. Whereas PIPLC interferes with pronephric duct cell migration but not somite fissure formation, trypsin in the presence of calcium does precisely the opposite. The differential sensitivities of these two morphogenetic movements to the different enzyme treatments provide further evidence for a degree of selectivity in the action of PIPLC on morphogenetic systems.

In summary, the ability of PIPLC to inhibit pronephric duct cell migration provides strong evidence that at least one species of PI-G anchored cell surface molecule, possibly alkaline phosphatase, plays an active role in the directed migration of the pronephric duct. It has been speculated that the PI-G linkage is useful in situations in which a cell surface protein is needed for a particular task but then must be rapidly (and specifically) removed from the surface (Low, 1987; Low & Saltiel, 1988; Cross, 1987). It is reasonable to conjecture that upon completion of a particular morphogenetic task a molecule with cell guidance activity might need to be removed rapidly from cells to avoid interference with subsequent steps of development. Several other PI-G anchored molecules implicated in cell adhesion, including a PI-G linked form of N-CAM, have been reported (He et al. 1986; Low, 1987). We speculate that the PI-G linkage might be commonly used to anchor other molecules involved in directing morphogenesis.

We thank the Indiana University Axolotl Colony for providing us with axolotl embryos, and Dr Martin Low for providing us with PIPLC. Part of the initial development and characterization of the Hypa beads was conducted by Mr Edward Niestat in partial fulfillment of the requirements for the degree of Bachelor of Arts at Princeton University. We thank Elaine Lenk for technical assistance with scanning electron microscopy. Supported by National Cancer Institute Grant 5R01 CA 13605.