Endocytosis without clathrin.

Internalization of membrane, fluid and receptor-bound ligands into cells occurs by at least two endocytic mechanisms. One is dependent on clathrin and responsible for concentrative uptake of growth factors and other ligands, whereas the other operates without clathrin. Clathrin-independent endocytosis, which might involve more than one mechanism, can contribute significantly to the total uptake of membrane and fluid in a cell. The properties and possible roles of clathrin-independent endocytosis are discussed in this article.

Are caveolae involved in clathrin-independent endocytosis?

In addition to endocytosing molecules via clathrin-coated pits, cells also internalize membrane and fluid by a clathrin-independent endocytic mechanism. In this article we search for the equivalent of clathrin-coated pits in clathrin-independent endocytosis, and discuss some pitfalls in the interpretation of electron micrographs. We also discuss how the early steps in clathrin-independent endocytosis might be analysed morphologically, and we argue that caveolae are not involved in clathrin-independent endocytosis.

The immunofluorescent era of membrane traffic.

In recent years immunofluorescence microscopy has been increasingly used to study membrane traffic. In this article seven electron microscopists, all with considerable experience in using light microscopy, take a critical look at the immunofluorescence approach and argue that results obtained with this method are often overinterpreted.

Membrane structure of nonactivated and activated human blood platelets as revealed by freeze-fracture: evidence for particle redistribution during platelet contraction.

The distribution of intramembrane particles of nonactivated and activated human blood platelets was studied by freeze-fracture under various experimental conditions to see whether morphological evidence for a structural coupling between the platelet actomyosin system and the fibrin network in a retracting clot could be established. Membrane particles were evenly distributed in nonactivated platelets; the total number (E + P faces) was approximately 1,500/micrometers 2 of membrane, and there were two to three times more particles present on the E face than on the P face. Transformation of discoid platelets to "spiny spheres" by cooling did not change the particle distribution. Platelet activation and aggregation by serum or ADP caused no change in membrane particle density or distribution. Particle distribution was not changed in Ca2+-activated platelets fixed immediately before fibrin formation, but after fibrin formation and during clot retraction, particles were sometimes most frequent on the P face and tended to form distinct clusters, and aggregates of E face pits were observed. Blood platelets contain contractile proteins that are distinct as filaments in platelets in retracting clots. We suggest that the redistribution of particles seen in activated platelets during clot retraction reflects the esablishment of mechanical transmembrane links between the platelet actomyosin system and the fibrin net. The P-face particle clusters may represent sites of force transmission between actin filaments bonded to the inside of the membrane and the fibrin network at the outside. Thus, whereas membrane particles may not be directly involved in the attachment of actin filaments to membranes, the transmission of the force of the contractile system to an exterior substrate apparently involves the intramembrane particles.

Pinocytosis in mouse L-fibroblasts: ultrastructural evidence for a direct membrane shuttle between the plasma membrane and the lysosomal compartment.

Mouse L-fibroblasts internalized large amounts of cationized ferritin (CF) by pinocytosis. Initially (60-90 s after addition of CF to cell monolayers at 37 degrees C), CF was found in vesicles measuring 100-400 nm (sectioned diameter) and as small clusters adhering to the inner aspect of the limiting membrane of a few large (greater than 600 nm) vacuoles. After 5-30 min, CF labeling of large vacuoles was pronounced and continuous. Moreover, 70-80% of all labeled structures were tiny (less than 100 nm) vesicles. However, the absolute frequency of tiny vesicles increased more than twofold from 5 min to 30 min. When the cells were incubated with CF for 30 min, then washed and further incubated for 3 h without CF, almost all CF was present in dense bodies (100-500 nm). When L-cells were first incubated with horseradish peroxidase (HRP), then washed and incubated with CF, double-labeled vacuoles were observed. Tiny vesicles also contained HRP-CF, and small HRP-CF patches were localized on the cell surface. Distinct labeling of stacked Golgi cisterns was not observed in any experiment. These observations suggest that the numerous tiny vesicles are not endocytic but rather pinch off from the large vacuoles and move towards the cell surface to fuse with the plasma membrane. Thus, ultrastructural evidence is provided in favor of a direct membrane shuttle between the plasma membrane and the lysosomal compartment.

Tight junctions in the choroid plexus epithelium. A freeze-fracture study including complementary replicas.

The tight junctions of the choroid plexus epithelium of rats were studied by freeze-fracture. In glutaraldehyde-fixed material, the junctions exhibited rows of aligned particles and short bars on P-faces, the E-faces showing grooves bearing relatively many particles. A particulate nature of the junctional strands could be established by using unfixed material. The mean values of junctional strands from the lateral, third, and fourth ventricles of Lewis rats were 7.5 +/- 2.6, 7.4 +/- 2.2, and 7.5 +/- 2.4; and of Sprague-Dawley rats 7.7 +/- 3.4, 7.4 +/- 2.3, and 7.3 +/- 1.6. Examination of complementary replicas (of fixed tissue) showed that discomtinuities are present in the junctional strands: 42.2 +/- 4.6% of the length of measured P-face ridges were discontinuities, and the total amount of complementary particles in E-face grooves constituted 17.8 +/- 4.4% of the total length of the grooves, thus approximately 25% of the junctional strands can be considered to be discontinuous. The average width of the discontinuities, when corrected for complementary particles in E-face grooves, was 7.7 +/- 4.5 nm. In control experiments with a "tighter" tight junction (small intestine), complementary replicas revealed that the junctional fibrils are rather continuous and that the very few particles in E-face grooves mostly filled out discontinuities in the P-face ridges. Approximately 5% of the strands were found to be discontinuous. These data support the notion that the presence of pores in the junctional strands of the choroid plexus epithelium may explain the high transepithelial conductance in a "leaky" epithelium having a high number of junctional strands. However, loss of junctional material during fracturing is also considered as an alternative explanation of the present results.

Vesicular transport of cationized ferritin by the epithelium of the rat choroid plexus.

We have studied the transport of ferritin that was internalized by coated micropinocytic vesicles at the apical surface of the choroid plexus epithelium in situ. After ventriculocisternal perfusion of native ferritin (NF) or cationized ferritin (CF), three routes followed by the tracers are revealed: (a) to lysosomes, (b) to cisternal compartments, and (c) to the basolateral cell surface. (a) NF is micropinocytosed to a very limited degree and appears in a few lysosomal elements whereas CF is taken up in large amounts and can be followed, via endocytic vacuoles and light multivesicular bodies, to dark multivesicular bodies and dense bodies. (b) Occasionally, CF particles are found in cisterns that may represent GERL or trans-Golgi elements, whereas stacked Golgi cisterns never contain CF. (c) Transepithelial vesicular transport of CF is distinctly revealed. The intercellular spaces of the epithelium, below the apical tight junctions, contain numerous clusters of CF particles, often associated with surface-connected, coated vesicles. Vesicles in the process of exocytosis of CF are also present at the basal epithelial surface, whereas connective tissue elements below the epithelium are unlabeled. Our conclusion is that fluid and solutes removed from the cerebrospinal fluid by endocytosis either become sequestered in the lysosomal apparatus of the choroidal epithelium or are transported to the basolateral surface. However, our results do not indicate any significant recycling via Golgi complexes of internalized apical cell membrane.

Serial-section analysis of coated pits and vesicles involved in adsorptive pinocytosis in cultured fibroblasts.

We have examined, by analyzing thin (15-20 nm) serial sections, whether coated pits involved in adsorptive pinocytosis in cultured fibroblasts give rise to free coated vesicles or represent permanently surface-associated structures from the neck of which uncoated receptosomes pinch off and carry ligand into the cell. Human skin fibroblasts and mouse L-929 fibroblasts were incubated with cationized ferritin (CF), a ligand known to bind to coated pit regions, at 37 degrees C before fixation. In thin sections, CF was found in coated vesicular profiles within the cytoplasm. Serial sections revealed that whereas many of these coated profiles communicated with the cell surface, thus representing pits, about 10% in L-cells and 36% in skin fibroblasts were actually free coated vesicles. Moreover, evidence for uncoated vesicular structures (receptosomes) budding off from the coated pits was not obtained. We therefore conclude that coated pits do pinch off from the plasma membrane to form free, coated vesicles (pinosomes).

A transferrin-like GPI-linked iron-binding protein in detergent-insoluble noncaveolar microdomains at the apical surface of fetal intestinal epithelial cells.

A GPI-anchored 80-kD protein was found to be the major component of detergent-insoluble complexes, prepared from fetal porcine small intestine, constituting about 25% of the total amount of protein. An antibody was raised to the 80-kD protein, and by immunogold electron microscopy of ultracryosections of mucosal tissue, the protein was localized to the apical surface of the enterocytes, whereas it was absent from the basolateral plasma membrane. Interestingly, it was mainly found in patches of flat or invaginated apical membrane domains rather than at the surface of microvilli. Caveolae were not found in association with these labeled microdomains. In addition, the 80-kD protein was seen in apical endocytic vacuoles and in tubulo-vesicular structures, suggesting that the apical microdomains are involved in endocytosis of the 80-kD protein. By its NH2-terminal amino acid sequence, iron-binding capacity and partial immunological cross-reactivity with serum transferrin, the 80-kD protein was shown to belong to the transferrin family, and it is probably homologous to melanotransferrin, a human melanoma-associated antigen. The 80-kD iron-binding protein was fully detergent-soluble immediately after synthesis and only became insoluble after gaining resistance to endo H, supporting a mechanism for exocytic delivery to the apical cell surface by way of detergent-insoluble glycolipid "rafts" that fuse with the plasmalemma at restricted sites devoid of microvilli.

The preendosomal compartment comprises distinct coated and noncoated endocytic vesicle populations.

The transfer of molecules from the cell surface to the early endosomes is mediated by preendosomal vesicles. These vesicles, which have pinched off completely from the plasma membrane but not yet fused with endosomes, form the earliest compartment along the endocytic route. Using a new assay to distinguish between free and cell surface connected vesicle profiles, we have characterized the preedosomal compartment ultrastructurally. Our basic experimental setup was labeling of the entire cell surface at 4 degrees C with Con A-gold, warming of the cells to 37 degrees C to allow endocytosis, followed by replacing incubation medium with fixative, all within either 30 or 60 s. Then the fixed cells were incubated with anti-Con A-HRP to distinguish truly free (gold labeled) endocytic vesicles from surface-connected structures. Finally, analysis of thin (20-30 nm) serial sections and quantification of vesicle diameters were carried out. Based on this approach it is shown that the preendosomal compartment comprises both clathrin-coated and non-coated endocytic vesicles with approximately the same frequency but with distinct diameter distributions, the average noncoated vesicle being smaller (95 nm) than the average coated one (110 nm). In parallel experiments, using an anti-transferrin receptor gold-conjugate as a specific marker for clathrin-dependent endocytosis it is also shown that uncoating of coated vesicles plays only a minor role for the total frequency of noncoated vesicles. Furthermore, after perturbation of clathrin-dependent endocytosis by potassium depletion where uptake of transferrin is blocked, noncoated endocytic vesicles with Con A-gold, but not coated vesicles, exist already after 30 and 60 s. Finally, it is shown that the existence of small, free vesicles in the short-time experiments cannot be ascribed to recycling from the early endosomes.

Molecules internalized by clathrin-independent endocytosis are delivered to endosomes containing transferrin receptors.

We have previously demonstrated that the preendosomal compartment in addition to clathrin-coated vesicles, comprises distinct nonclathrin coated endocytic vesicles mediating clathrin-independent endocytosis (Hansen, S. H., K. Sandvig, and B. van Deurs. 1991. J. Cell Biol. 113:731-741). Using K+ depletion in HEp-2 cells to block clathrin-dependent but not clathrin-independent endocytosis, we have now traced the intracellular routing of these nonclathrin coated vesicles to see whether molecules internalized by clathrin-independent endocytosis are delivered to a unique compartment or whether they reach the same early and late endosomes as encountered by molecules internalized with high efficiency through clathrin-coated pits and vesicles. We find that Con A-gold internalized by clathrin-independent endocytosis is delivered to endosomes containing transferrin receptors. After incubation of K(+)-depleted cells with Con A-gold for 15 min, approximately 75% of Con A-gold in endosomes is colocalized with transferrin receptors. Endosomes containing only Con A-gold may be accounted for either by depletion of existing endosomes for transferrin receptors or by de novo generation of endosomes. Cationized gold and BSA-gold internalized in K(+)-depleted cells are also delivered to endosomes containing transferrin receptors. h-lamp-1-enriched compartments are only reached occasionally within 30 min in K(+)-depleted as well as in control cells. Thus, preendosomal vesicles generated by clathrin-independent endocytosis do not fuse to any marked degree with late endocytic compartments. These data show that in HEp-2 cells, molecules endocytosed without clathrin are delivered to the same endosomes as reached by transferrin receptors internalized through clathrin-coated pits.

Clathrin and HA2 adaptors: effects of potassium depletion, hypertonic medium, and cytosol acidification.

The effects of methods known to perturb endocytosis from clathrin-coated pits on the localization of clathrin and HA2 adaptors in HEp-2 carcinoma cells have been studied by immunofluorescence and ultrastructural immunogold microscopy, using internalization of transferrin as a functional assay. Potassium depletion, as well as incubation in hypertonic medium, remove membrane-associated clathrin lattices: flat clathrin lattices and coated pits from the plasma membrane, and clathrin-coated vesicles from the cytoplasm, as well as those budding from the TGN. In contrast, immunofluorescence microscopy using antibodies specific for the alpha- and beta-adaptins, respectively, and immunogold labeling of cryosections with anti-alpha-adaptin antibodies shows that under these conditions HA2 adaptors are aggregated at the plasma membrane to the same extent as in control cells. After reconstitution with isotonic K(+)-containing medium, adaptor aggregates and clathrin lattices colocalize at the plasma membrane as normally and internalization of transferrin resumes. Acidification of the cytosol affects neither clathrin nor HA2 adaptors as studied by immunofluorescence microscopy. However, quantitative ultrastructural observations reveal that acidification of the cytosol results in formation of heterogeneously sized and in average smaller clathrin-coated pits at the plasma membrane and buds on the TGN. Collectively, our observations indicate that the methods to perturb formation of clathrin-coated vesicles act by different mechanisms: acidification of the cytosol by affecting clathrin-coated membrane domains in a way that interferes with budding of clathrin-coated vesicles from the plasma membrane as well as from the TGN; potassium depletion and incubation in hypertonic medium by preventing clathrin and adaptors from interacting. Furthermore our observations show that adaptor aggregates can exist at the plasma membrane independent of clathrin lattices and raise the possibility that adaptor aggregates can form nucleation sites for clathrin lattices.

Acidification of the cytosol inhibits endocytosis from coated pits.

Acidification of the cytosol of a number of different cell lines strongly reduced the endocytic uptake of transferrin and epidermal growth factor. The number of transferrin binding sites at the cell surface was increased in acidified cells. Electron microscopic studies showed that the number of coated pits at the cell surface was not reduced in cells with acidified cytosol. Experiments with transferrin-horseradish peroxidase conjugates and a monoclonal anti-transferrin receptor antibody demonstrated that transferrin receptors were present in approximately 75% of the coated pits both in control cells and in cells with acidified cytosol. The data therefore indicate that the reason for the reduced endocytic uptake of transferrin at internal pH less than 6.5 is an inhibition of the pinching off of coated vesicles. In contrast, acidification of the cytosol had only little effect on the uptake of ricin and the fluid phase marker lucifer yellow. Ricin endocytosed by cells with acidified cytosol exhibited full toxic effect on the cells. Although the pathway of this uptake in acidified cells remains uncertain, some coated pits may still be involved. However, the data are also consistent with the possibility that an alternative endocytic pathway involving smooth (uncoated) pits exists.

Endocytosis and intracellular transport of the glycolipid-binding ligand Shiga toxin in polarized MDCK cells.

The glycolipid-binding cytotoxin produced by Shigella dysenteriae 1, Shiga toxin, binds to MDCK cells (strain 1) only after treatment with short-chain fatty acids like butyric acid or with the tumor promoter 12-O-tetradecanoylphorbol 13-acetate. The induced binding sites were found to be functional with respect to endocytosis and translocation of toxin to the cytosol. Glycolipids that bind Shiga toxin appeared at both the apical and the basolateral surface of polarized MDCK cells grown on filters, and Shiga toxin was found to be endocytosed from both sides of the cells. This was demonstrated by EM of cells incubated with Shiga-HRP and by subcellular fractionation of cells incubated with 125I-labeled Shiga toxin. The data indicated that toxin molecules are endocytosed from coated pits, and that some internalized Shiga toxin is transported to the Golgi apparatus. Fractionation of polarized cells incubated with 125I-Shiga toxin showed that the transport of toxin to the Golgi apparatus was equally efficient from both poles of the cells. After 1-h incubation at 37 degrees C approximately 10% of the internalized toxin was found in the Golgi fractions. The results thus suggest that glycolipids can be efficiently transported to the Golgi apparatus from both sides of polarized MDCK cell monolayers.

Ricin transport in brefeldin A-treated cells: correlation between Golgi structure and toxic effect.

Whereas brefeldin A (BFA) protected a number of cell lines against the protein toxin ricin, two of the cell lines tested were not protected but rather sensitized to ricin by BFA. EM studies revealed that upon addition of BFA the Golgi stacks in cells which were protected against the toxin rapidly transformed into a characteristic tubulo-vesicular reticulum connected to the endoplasmic reticulum, and subcellular fractionation experiments showed that galactosyl transferase disappeared from the Golgi fractions where it was normally located. EM and subcellular fractionation also indicated that in contrast to the Golgi stacks, the trans-Golgi network (TGN) remained intact and that internalized ricin was still localized in the TGN both when BFA was added before and after the toxin. Thus, BFA does not prevent fusion of ricin-containing vesicles with the TGN, and unlike resident proteins in Golgi stacks, ricin is not transported back to ER upon treatment of cells with BFA. Two kidney epithelial cell lines, MDCK and PtK2, were not protected against ricin by BFA, and EM studies of MDCK cells revealed that BFA did not alter the morphology of the Golgi complex in these cells. Also, subcellular fractionation revealed that, in contrast to the other cell types tested, the localization of galactosyl transferase in the gradients was not affected by BFA treatment. The data show that there is a correlation between BFA-induced disassembly of the Golgi stacks and protection against ricin, and they demonstrate that the structural organization of the Golgi apparatus is affected by BFA to different extents in various cell lines.

Effects of brefeldin A on endocytosis, transcytosis and transport to the Golgi complex in polarized MDCK cells.

We have studied the effects of brefeldin A (BFA) on endocytosis and intracellular traffic in polarized MDCK cells by using the galactose-binding protein toxin ricin as a membrane marker and HRP as a marker of fluid phase transport. We found that BFA treatment rapidly increased apical endocytosis of both ricin and HRP, whereas basolateral endocytosis was unaffected, as was endocytosis of HRP in the poorly polarized carcinoma cell lines HEp-2 and T47D. Tubular endosomes were induced by BFA both apically and basolaterally in some MDCK cells, comparable with those seen in HEp-2 and T47D cells. In addition, in MDCK cells, BFA induced formation of small (< 300 nm) vesicles, labeled both after apical and basolateral uptake of HRP, as well as some very large (> 700 nm) vacuoles, which were only labeled when HRP was present in the apical medium. In contrast, neither in MDCK nor in HEp-2 or T47D cells, did BFA have any effect on lysosomal morphology. Moreover, transcytosis in the basolateral-apical direction was stimulated both for HRP and ricin. Other vesicular transport routes were less affected or unaffected by BFA treatment. Two closely related structural analogues of BFA (B16 and B21), unable to produce the changes in Golgi and endosomal morphology seen after BFA treatment in a number of different cell lines, were also unable to mimic the effects of BFA on MDCK cells.

Endocytosis from coated pits of Shiga toxin: a glycolipid-binding protein from Shigella dysenteriae 1.

Evidence is presented that endocytosis is involved in the transport to the cytosol of the cytotoxin from Shigella dysenteriae 1, Shiga toxin, which acts by removal of a single adenine residue in 28-S ribosomal RNA. Inhibition of endocytosis by ATP depletion of the cells prevented toxin uptake. Exposure of HeLa S3 and Vero cells to toxin at low extracellular pH, where translocation to the cytosol, but not endocytosis is inhibited, allowed the toxin to accumulate in a compartment where it was protected against antibodies to the toxin. Upon transfer of the cells to normal medium endocytosed toxin entered the cytosol. Electron microscopical studies of cells exposed at 0 degrees C to a toxin-horseradish peroxidase (HRP) conjugate, or to unconjugated toxin followed by horse antitoxin antibodies and then protein G-gold, revealed that the Shiga toxin binding sites were randomly distributed on the cell surface, without any preference to, for example, coated pits. In contrast, when cells were exposed to toxin at 37 degrees C, the binding sites were preferentially localized in coated pits. The Shiga-HRP conjugate was also seen in endosomes, lysosomes, and in the Golgi region. Endocytosis by the coated pit/coated vesicle pathway was selectively inhibited by acidification of the cytosol. Under these conditions, both the uptake of toxin-HRP conjugates and intoxication of the cells were inhibited. Evidence from the literature as well as our own results suggest that Shiga toxin binding sites are glycolipids. Thus, Shiga toxin appears to be the first example of a lipid-binding ligand that is endocytosed from coated pits.

Estimation of the amount of internalized ricin that reaches the trans-Golgi network.

We have used a protocol for internalization of ricin, a ligand binding to plasma membrane glycoproteins and glycolipids with terminal galactosyl residues, and infection with the vesicular stomatitis virus ts 045 mutant in BHK-21 cells to determine whether internalized plasma membrane molecules tagged by ricin reach distinct compartments of the biosynthetic-exocytic pathway. At 39.5 degrees C newly synthesized G protein of ts 045 was largely prevented from leaving the endoplasmic reticulum. At the same temperature ricin was endocytosed and reached, in addition to endosomes and lysosomes, elements of the Golgi complex. When the temperature was lowered to 19.5 degrees C, no more ricin was delivered to the Golgi complex, but now G protein accumulated in the Golgi stacks and the trans-Golgi network (TGN). Double-labeling immunogold cytochemistry on ultracryosections was used to detect G protein and ricin simultaneously. These data, combined with stereological and biochemical methods, showed that approximately 5% of the total amount of ricin within the cells, corresponding to 6-8 X 10(4) molecules per cell, colocalized with G protein in the Golgi complex after 60 min at 39.5 degrees C. Of this amount approximately 70-80% was present in the TGN. Since most of the ricin molecules remain bound to their binding sites at the low pH prevailing in compartments of the endocytic pathway, the results indicate that a fraction of the internalized plasma membrane molecules with terminal galactose are not recycled directly from endosomes or delivered to lysosomes, but are routed to the Golgi complex. Also, the results presented here, in combination with other recent studies on ricin internalization, suggest that translocation of the toxic ricin A-chain to the cytosol occurs in the TGN.

Analysis of GLUT4 distribution in whole skeletal muscle fibers: identification of distinct storage compartments that are recruited by insulin and muscle contractions.

The effects of insulin stimulation and muscle contractions on the subcellular distribution of GLUT4 in skeletal muscle have been studied on a preparation of single whole fibers from the rat soleus. The fibers were labeled for GLUT4 by a preembedding technique and observed as whole mounts by immunofluorescence microscopy, or after sectioning, by immunogold electron microscopy. The advantage of this preparation for cells of the size of muscle fibers is that it provides global views of the staining from one end of a fiber to the other and from one side to the other through the core of the fiber. In addition, the labeling efficiency is much higher than can be obtained with ultracryosections. In nonstimulated fibers, GLUT4 is excluded from the plasma membrane and T tubules. It is distributed throughout the muscle fibers with approximately 23% associated with large structures including multivesicular endosomes located in the TGN region, and 77% with small tubulovesicular structures. The two stimuli cause translocation of GLUT4 to both plasma membrane and T tubules. Quantitation of the immunogold electron microscopy shows that the effects of insulin and contraction are additive and that each stimulus recruits GLUT4 from both large and small depots. Immunofluorescence double labeling for GLUT4 and transferrin receptor (TfR) shows that the small depots can be further subdivided into TfR-positive and TfR-negative elements. Interestingly, we observe that colocalization of TfR and GLUT4 is increased by insulin and decreased by contractions. These results, supported by subcellular fractionation experiments, suggest that TfR-positive depots are only recruited by contractions. We do not find evidence for stimulation-induced unmasking of resident surface membrane GLUT4 transporters or for dilation of the T tubule system (Wang, W., P.A. Hansen, B.A. Marshall, J.O. Holloszy, and M. Mueckler. 1996. J. Cell Biol. 135:415-430).

Routing of internalized ricin and ricin conjugates to the Golgi complex.

Receptor-mediated endocytosis and intracellular routing of native ricin, and of ricin conjugated to colloidal gold (Ri-Au) and to horseradish peroxidase (Ri-HRP), have been studied in cultured MCF-7 and Vero cells by electron microscopical techniques including serial section analysis. Both native ricin, as demonstrated by immunoperoxidase cytochemistry, and the ricin conjugates were internalized via a common coated pit-coated vesicle pathway to reach vacuolar and tubulo-vesicular portions of the endosomal system. In addition, native ricin and a purified monovalent fraction of Ri-HRP reached distinct Golgi cisterns, whereas Ri-Au and polyvalent Ri-HRP did not. The results delineate intracellular routing of native ricin and compare it with the routing of different ricin conjugates. Moreover, our study shows that conjugates of a particular ligand (ricin) and various probes (e.g., gold and peroxidase), may be handled differently by cells. Sorting apparently takes place in the endosomal system, allowing some but not other molecules to reach Golgi elements. This sorting seems to depend on the valency of the ricin conjugate.