Endocytosis, Recycling, and Lysosomal Delivery of Brush Border Hydrolases in Cultured Human Intestinal Epithelial Cells (Caco-2)”

Lysosomes of intestinal epithelial cells in uivo and in culture display strong immunoreactivity with mono- clonal antibodies against various brush border en- zymes as visualized by immunoelectron microscopy. Novel subcellular fractionation procedures were de- veloped to study, by the pulse-chase technique and by internalization assays, the pathway along which two microvillar hydrolases, sucrase-isomaltase and dipeptidylpeptidase IV, are transported to lysosomes in the differentiated colon adenocarcinoma cell line Caco-2. 7-9% of metabolically labeled sucrase-isomaltase of dipeptidylpeptidase IV were present in lysosomes after 7-8 h of chase as intact complex-glycosylated mole- cules. Appearance of these enzymes in lysosomes was biphasic.

Endocytosis studies with radioiodinated antienzyme monoclonal antibodies (monovalent antigenbinding fragments) and by means of cell surface iodination revealed only slow transport of the enzymes to lysosomes at a low level. However, both enzymes were internalized with different efficiencies and recycled to the cell surface via endosomes. These results suggest that in Caco-2 cells a significant amount of newly synthesized sucrase-isomaltase and dipeptidylpeptidase IV is directly imported into lysosomes bypassing the brush border membrane.
Newly synthesized proteins that pass through the Golgi apparatus have been classified into three major groups characterized by their exit pathways from the trans-Golgi network, plasma membrane and constitutively secreted proteins, secretory proteins that are released from the cell in a regulated fashion, and lysosomal enzymes (Griffiths and Simons, 1986). In kidney epithelial cells, but not in hepatocytes (Bartles et al., 1987;Bartles and Hubbard, 1988), the first pathway is further subdivided into an apical and a basolateral route that split at the exit site of the Golgi apparatus (Matlin and Simons, 1983;Misek et al., 1984, Griffiths et al., 1985Rindler et al., 1985;Wandinger-Ness and Simons, 1988). It is possible that this general framework of trans-Golgi sorting needs to be modulated in order to accommodate multiple localization patterns observed for a number of different proteins including the cell surface localization of the mannose 6-phosphate receptor (Geuze et al., 1984(Geuze et al., ,1985Griffiths et al., 1988) and of a lysosomal membrane glycoprotein (Lippincott-*This work was supported by Grant 3622087 from the Swiss National Science Foundation and by a grant from the Ciba Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. q To whom correspondence should be addressed. Schwartz and Fambrough, 1987) as well as the brush border localization of lysosomal acid a-glucosidase . Presently, it is unclear if these localizations are due to missorting of a subset of a given protein or whether there is a dynamic equilibrium between cell surface and lysosomes as proposed by various authors (Kornfeld, 1987;Lippincott-Schwartz et al., 1987).
An interesting case of unexpected protein localization concerns the apparent presence of intestinal brush border enzymes in lysosomes of intestinal epithelial cells as visualized by immunoelectron microscopy (Fransen et al., 1985;Sips et al., 1985;Lorenzsonn et al., 1987;Fransen et al., 1989). To date at least two brush border enzymes, sucrase-isomaltase and dipeptidylpeptidase IV (DPPIV),' have been localized in this organelle. The lysosomal immunoreactivity is not due to carbohydrate epitopes, which may be common to nonrelated glycoproteins, since most studies were performed with wellcharacterized protein epitope-specific mAbs. On the basis of rather indirect evidence it has been speculated that a fraction of newly synthesized brush border enzymes may be directly imported into lysosomes from the Golgi apparatus (Blok et al., 1984).
In the present study we have addressed the question of whether brush border enzymes in intestinal lysosomes originate from the brush border membrane and are transported to this organelle by means of endocytosis, or whether they have bypassed the apical membrane on their way to lysosomes. We approached this question by studying the trafficking of brush border sucrase-isomaltase, a representative disaccharidase, and DPPIV, a representative peptidase, in the differentiated colon adenocarcinoma cell line Caco-2 (Pinto et al., 1983;Zweibaum et al., 1988).
Disaccharidases and peptidases are the major glycoproteins of the small intestinal brush border membrane (Kenny and Maroux, 1982;Semenza, 1986;Noren et al., 1986;Hauri, 1988) some of which are expressed in Caco-2 cells (Zweibaum et al., 1989;Hauri, 1988). In these cells the biogenesis and intracellular transport of sucrase-isomaltase and DPPIV have been studied in detail. The two enzymes were found to mature at different rates (Hauri et al., 1985a). This asynchronism is due to at least two rate-limiting steps along the rough endoplasmic reticulum to trans-Golgi pathway as established by subcellular fractionation (Stieger et al., 1988). To study the transport of brush border enzymes to lysosomes, we have now developed an isolation procedure for lysosomes of Caco-2 cells. Using this method in conjunction with metabolic and cell surface labeling experiments, we provide evidence that a significant amount of newly synthesized sucrase-isomaltase and DPPIV is directly targeted to lysosomes and bypasses the brush border membrane. Furthermore, evidence is presented for endocytosis and recycling of brush border hydrolases.

MATERIALS AND METHODS
Cell Culture and Labeling with r'S]Methionine-Caco-2 cells, kindly provided by Dr. A. Zweibaum (Paris), were grown in Optilux Petri dishes as described (Pinto et al., 1983;Hauri et al., 1985a) (Gorr et al., 1988). This mAb does not recognize any protein in Caco-2 cells. Furthermore, antibody G1/139 against a lysosomal membrane glycoprotein (Schweizer et aZ., 1988), and HBB3/456 a new mAb against DPPIV were used. The latter antibody was derived from the same fusion as HBB3/775 and was found to precipitate 52% of the DPPIV activity of solubilized Caco-2 membranes (using glycyl-L-proline-p-nitroanilide-p-tosylate as substrate). The antigen recognized by HBB3/456 has the same molecular weight as that of the HBB3/775 antigen. SDS-PAGE was performed as described (Hauri et aZ., 1985a) using 7.5%-gels.
[?S]Methioninelabeled proteins were visualized by fluorography using En3Hance (Du Pont-New England Nuclear). Bands on fluorograms were quantified by means of a Camag LTC Scanner II connected on a Camag SP4290 integrator.
Isolation of a Fraction Enriched in Lysosomal Membranes-All the steps of the isolation procedure were carried out at 4 "C unless indicated otherwise. Caco-2 cell monolayers were washed once with 0.9% (w/v) NaCl and once with 250 mM sucrose, 10 mM triethanolamine-acetic acid, pH 7.4 (buffer A). The cells were then scraped from the dish, suspended in 2 ml/lOO-mm dish buffer A, and centrifuged for 5 min at 560 g.,. The cell pellet was gently resuspended in 2 ml of 250 mM sucrose, 1 mM Na*EDTA, 10 mM triethanolamine-acetic acid, pH 6.5 (buffer B), and centrifuged as above. The cells were then resuspended in 3 ml of buffer B and homogenized by passing them 10 times through a ball-bearing homogenizer (Balch and Rothman. 1985) with a clearance of 20 irn. The resulting homogenates were pooled and centrifuged for 10 min at 2,000 rpm (370 &) in an SS34 rotor (Sorvall Instriments Division). The supernatant was brought to exactly 30 ml with buffer B, and 4.66 ml stock isoosmotic Percoll (density of Percoll = 1.129, initial density = 1.048) were added. The Percoll gradient (see Fig. 1) was centrifuged for 41 min at 20,000 rpm (36,900 g.,) in an RC 2B centrifuge (Sorvall Instruments Division) using the same rotor as above. 2 ml from the bottom of the gradient were harvested and processed as follows: 2 g of this fraction were mixed with 2 g of 60% (w/w) Metrizamide in 1 mM Na2EDTA, 10 mM triethanolamine-acetic acid, pH 6.5 (buffer C) in a centrifuge tube and overlayed with 2.5 ml each of 21.5 and 14.5% (w/w) Metrizamide in buffer C. Finally, the tube was filled with buffer B. The gradient was run for 5 h at 23,000 rpm (70,600 g.") at 8 "C in a TST41.14 rotor (Kontron Elektronik GmbH, Zurich at 37 "C in PBS containing 2 mM EDTA and 25 mM /3mercaptoethanol. After 10 min papain (Merck, Darmstadt, Germany) was added to a final concentration of 40 pg/ml. The digestion was stopped after 1 h at 37 "C by the addition of iodoacetamide (30 mM final concentration), and the sample was dialyzed against 10 mM NaH2P0,, pH 7.6. Fab fragments were purified by DEAE-52 column chromatography using a gradient with limit buffer of 10 mM Na-H2P04, 1 M NaCl, pH 7.6. Purity of the Fab fraction was tested by SDS-PAGE (10% eels) that were stained with silver (Merril et al.. 1984). After clialysiHag:inst PBS the fragments were stoied at -20 'C: Fab fragments were iodinated by the chloramine-T method (Hunter and Greenwood, 1962). For the labeling of 5 pg of protein 0.5 mCi Na [lz51] was used, and free iodine was removed by gel filtration on Sephadex G25. The labeled fragments were stored at 4 "C and used within 2 weeks. Cell Surface Labeling with lz5Z-Fab Fragments-Mini-Marbrook chambers containing the filter-grown Caco-2 cells were transferred to 6-well plates and labeled at 4 "C as follows.
The cells were rinsed once in PBS containing 0.1% bovine serum albumin and were incubated in the same buffer. After 10 min the chambers were transferred to a fresh 6-well plate containing Caco-2 medium.
The '*'I-Fab fragments (5 x lo6 cpm in 350 ~1 of medium) were added to the apical medium.
After 1 h the chamber was rinsed twice with PBS, disassembled, and washed three times for 3 min each with PBS containing 0.1% bovine serum albumin.
Depending on the experiment the labeled cells were either harvested directly or incubated in Caco-2 medium at 37 "C. In some experiments the cell surface label was removed by a low pH treatment.
For this purpose the cells were rinsed twice with 100 DIM glycin, 250 DIM sucrose, pH 2.0, followed by three 20-min incubations in the same buffer. By this procedure more than 98% of cell surface radioactivity was removable from cells that were labeled and acid-treated at 4 "C. Thereafter, the cells were washed once more with the acidic buffer and twice with PBS. After incubating the cells twice in PBS for 10 min, they were harvested in buffer A for subcellular fractionation or in PBS for direct determination of radioactivity in a y-counter (GAMMAmatic, propionate; Pierce, The Netherlands), a membrane impermeable form of the Bolton-Hunter reagent (Thompson et al., 1987) was used for cell surface iodination.
Iodination of the reagent and the labeling of the cell surface was done according to Bretscher and Lutter (1988). Brieflv. to 15 ~1 of Hz0 and 5 ul of Na1Yl (0.5 mCi) was added 1.~1 of 3 -4 NaCl; 2 ~1 oi 1 M K&POh, pH 7.b, 1.5 ~1 of stock reagent (1 mg/ml Sulfo-SHPP in dimethyl sulfoxide) and 3 rl of 5 mg/ml chloramine T and this mix was kept on ice for 15 min. To this was added 3 pl of a solution of 700 mM hydroxyphenylacetic acid, pH 7.0, containing 0.1 M NaI. After 3 min on ice the labeled reagent was diluted with 250 ~1 of 10 mM triethanolamine/HCl, pH 7.4, containing 250 mM sucrose and 2 mM Call,. Before labeling, the cells grown in mini-Marbrook chambers were washed three times with Dulbecco's PBS for 5 min each. by adapting the glutathione stripping procedure described by Bretscher and Lutter (1988) to cells grown in mini-Marbrook chambers.
The chambers were disassembled and the filters were transfered to a 6-well plate containing 6 ml/well glutathione solution at 4 "C. Preparation of glutathione solution: 155 mg glutathione (reduced form, Fluka, Switzerland) was dissolved in 7.25 ml of water and 1 ml of IO-times concentrated TBS. Immediately before use 750 ~1 of 1 N NaOH and then 1 ml of 10% BSA were added to the dissolved glutathione.
The 6-well plate was placed on a horizontal shaker for 20 min at 4 "C. Then the cells were transferred to fresh glutathione solution and shaken for an additional 20 min. This step was repeated once, and the cells were rinsed twice in PBS containing 0.1% BSA and twice in Caco-2 medium.
Cell surface biotin was detected by the use of streptavidin (Pierce, The Netherlands) which was iodinated as described for Fab fragments. The cells were labeled at 4 "C in 2 ml of Caco-2 medium containing 10' cpm iz51-streptavidin. After 5 min on ice, the sample was centrifuged for 5 min at 13,400 g.,. The resulting pellet was resuspended at 37 "C in 10% gelatin dissolved in PBS and immediately centrifuged as above. The supernatant was removed and the sample was postfixed with 1% paraformaldehyde in 100 mM phosphate buffer. Cryosectioning and immunogold-labeling was done as described (Fransen et al., 1985;Schweizer et al., 1988).

Appearance of Metabolically Labeled Sucrase-Isomaltase and DPPZV in Lysosomes-Recent
immunocytochemical labeling studies at the ultrastructural level with mAbs have suggested that lysosomes of Caco-2 cells may harbor sucrase-isomaltase and DPPIV (Fransen et al., 1988b) analogous to lysosomes of small-intestinal enterocytes in uiuo (Fransen et al., 1985;Lorenzsonn et al., 1987;Hauri et al., 1985b;Sips et al., 1985). In order to investigate whether these cytochemical signals are indeed due to intact polypeptides of sucrase-isomaltase and DPPIV and to study the transport of these hydrolases to lysosomes, we developed an isolation procedure for Caco-2 lysosomes (Fig. 1)   this method is identical to the preparation of Golgi-derived membranes (Stieger et al., 1988). Percoll gradient centrifugation already led to a good separation of lysosomes from membranes derived from other organelles as deduced from marker enzyme measurements (not shown). To achieve a higher purity of lysosomes the bottom fraction of the Percoll gradient was fractionated on a Metrizamide gradient (Fig. 1). The enzymatic characterization of the final fraction, designated "lysosomal fraction" (LII), is given in Tables I and II. This fraction was 24fold enriched in lysosomal glucosaminidase activity while enzymatic activities for other cellular membrane markers were not enriched. An ultrastructural analysis of the lysosomal fraction revealed dense organelles resembling lysosomes (Fig. 2). The lysosomal origin of these organelles was confirmed by immunoelectron microscopy using a mAb against the lysosomal enzyme a-glucosidase ( Fig.  2A). This antibody is specific for the intermediate and mature forms of cu-glucosidase but does not react with the IlO-kDa precursor form . The lysosome-like structures also showed immunoreactivity with a mAb against DPPIV (Fig. 2B) and a mAb against sucrase-isomaltase (not shown). The signal with the latter antibody was weak but clearly above background. A pulse-chase protocol was used to study the arrival in the lysosomal fraction of newly synthesized sucrase-isomaltase and DPPIV. A typical fluorograph of such an experiment is given in Fig. 3 which shows that the molecular mass of the hydrolases in lysosomes is indistinguishable from the complex-glycosylated enzyme forms of the homogenate suggesting that the visible polypeptides represent intact enzymes. The fact that no degradation products appeared on the gels may be due to an inability of our mAbs to recognize proteolytic fragments of the enzymes. The kinetics of appearance are drawn in Fig. 4. The kinetic behavior of the two digestive hydrolases is relatively similar during the first 10 h of chase. DPPIV showed a peak after 7 h and sucrase-isomaltase after 8 h. This delayed appearance is likely to be due to the  asynchronous transport of the two enzymes to and through the Golgi apparatus (Stieger et al., 1988). The maximal fraction of newly synthesized enzyme in lysosomes was about 9% for sucrase-isomaltase and 7% for DPPIV (values are corrected for the yield of the fraction). It is unlikely that this result is due to cross-contamination of the lysosomal fraction with membranes of other organelles for two reasons. First, the enzymatic data of the fraction LII (Tables I and III) do not indicate such cross-contamination problems and second, the kinetics of appearance of DPPIV and sucrase-isomaltase in lysosomes is strikingly different from that in other organelles including the Golgi apparatus and the brush border membrane. (Stieger et al., 1988).' The finding that both enzymes arrive in lysosomes later than in the brush border membrane would be compatible with a mechanism whereby the proteins are first exported to the cell surface followed by endocytosis. Interestingly, the profiles of metabolically labeled sucrase-isomaltase and DPPIV in LII are biphasic, in particular for DPPIV, suggesting that the hydrolases enter lysosomes in two waves. In subsequent experiments we tested the assumption that the first. wave reflects a pathway which bypasses the brush border membrane and that the second wave originates from endocytosis.

Transport of Brush Border Enzymes to Lysosomes
Endocytosis of Sucrase-Isomaltase and DPPIV-Assuming that all of the newly synthesized sucrase-isomaltase or DPPIV would first be inserted into the brush border and subsequently internalized to lysosomes one would expect the maximal appearance in lysosomes of cell surface-labeled brush border enzymes (that is the maximal lysosomal radioactivity relative to the homogenate) to be in the same order of magnitude as that found after metabolic labeling with [35S]methionine.
To study a possible endocytosis of brush border enzymes, we established a cell surface labeling assay. In this assay Caco-2 cells were labeled at 4 "C with lz51-Fab fragments of antienzyme mAbs. After removal of the unbound antibodies the cells were returned to the 37 "C incubator and the timedependent uptake of radioactivity into the cell was measured. The use of Fab fragments instead of the divalent intact immunoglobulin G is important since cross-linking of brush border enzymes by divalent antibodies may induce endocytosis (Louvard et al., 1980). The availability of a highly specific cell surface labeling assay is crucial for endocytosis experiments. Additional labeling of unrelated antigens that are static in the brush border membrane may mask a low rate of endocytosis or, in the opposite case, nonspecifically labeled components, if efficiently endocytosed, may lead to an artificially high apparent rate of endocytosis.
In a first control experiment, Caco-2 cells were separately labeled with the Fab fragments of HBB2/614 against sucrase-isomaltase, HBB3/ 775 against DPPIV and CP1/126 against an antigen of rat proximal colon (Gorr et al., 1988). The latter antibody does not react with any antigen of Caco-2 cells and was used as a negative control. The incubations with anti-sucrase-isomaltase-Fab or anti-DPPIV-Fab led to a high cell-associated radioactivity whereas the control antibody produced a signal only minimally above the background of the y-counter (data not shown). This indicates high specific binding of the antihydrolase antibodies.
However, different mAbs may respond differently to the limited proteolysis conditions of the Fab preparation, i.e. some may become sticky due to partial denaturation, and hence the above result may simply reflect different degrees of stickiness. To rule out this possibility we performed competition experiments.
The cells were incubated with either IgG HBB2/614 or HBB3/775 in excess, or without antibody, before labeling with either Fab HBB2/614  Caco-2 cells were labeled from the apical side at 4 "C with lZ51-Fab fragments specific for DPPIV (x) or sucrase-isomaltase (0). The labeled cells were incubated for different time intervals at 37 'C and the ""I-Fab fragments remaining at the cell surface were removed by a pH 2.0 treatment. The cells were harvested and cell-associated radioactivity was determined. Control cultures which were neither incubated at 37 "C nor washed with low pH-buffer were set to 100%. Note that the scales are different for the two enzymes.
The total rate of endocytosis was measured in experiments that are analogous to a metabolic pulse-chase protocol. The cells were labeled with lZ51-Fab fragments at 4 "C and were then cultured at 37 "C for different time intervals. Thereafter, the cells were cooled down to 4 "C and the Fab fragments specifically bound to the cell surface were removed by a low pH treatment.
In this assay, acid-resistant radioactivity reflects internalized enzyme. The results of these experiments showed internalization of DPPIV and sucrase-isomaltase (Fig.  6). The shape of the two curves are quite similar and level off after 60-90 min at 37 "C. Endocytosis of DPPIV reaches a plateau after 60 min. It is worth noting, however, that the magnitude of endocytosed radioactivity was different for the two enzymes. While the plateau for DPPIV was at 18% of total radioactively labeled DPPIV, the endocytosis of sucraseisomaltase did not exceed 2% of totally labeled sucrase-isomaltase even after 6 h at 37 "C. The leveling off of the curves indicates that, if at all, only a small fraction of endocytosed sucrase-isomaltase and DPPIV are transported all the way to the lysosomes. Rather both enzymes appear to recycle to the cell surface (see below).

Delivery of Hydrolases from Brush Border to Lysosomes-
To measure directly the transport of sucrase-isomaltase and DPPIV from the plasma membrane to lysosomes the cell surface labeling approach was combined with subcellular fractionation.
Cell surface sucrase-isomaltase or DPPIV were labeled with 1251-Fab at 4 "C! and after various time intervals at 37 "C the cells were subjected to the isolation of lysosomes. These experiments showed a low rate of enzyme transport to lysosomes (Fig. 7) which was l-2% after 3 h and 2.5-4.5% after 18 h (values are corrected for the yield of the preparation). Since metabolically labeled sucrase-isomaltase and DPPIV need about 7-8 h for their maximal appearance in the lysosomal fraction (Fig. 4) and 5 h (sucrase-isomaltase) or 3 h (DPPIV) for maximal appearance in the brush border (Stieger et al., 1988) the important time window of the endocytosis experiment is about 3 h. At that time the percentage of endocytosed DPPIV and sucrase-isomaltase in the lysoso-ma1 fraction is way below the 7-9% (values are corrected for the yield of the preparation) of metabolically labeled sucraseisomaltase and DPPIV in that fraction (see Fig. 4).
This low delivery to lysosomes of iodinated Fab fragments could be due to a rapid degradation of the antibody fragments in lysosomes and loss of the iodinated peptides or tyrosine residues from lysosomes. However, in neither the incubation medium nor the cells were we able to detect trichloroacetic acid-soluble radioactivity even after incubations of up to 6 h at 37 "C. Another potential problem with this assay concerns the possibility that there might be a rapid release of Fab fragments into the medium upon warming up the cells. Indeed there is a low but slow release into the medium which levels off after 1 h at 37 "C. Maximal release into the medium was 15% for sucrase-isomaltase and 18% for DPPIV. However, this does not critically affect the results of Fig. 7 since the radioactivity in lysosomes was compared with the counts obtained in the corresponding homogenate. Nevertheless, to circumvent these potential problems we used an alternative method. Cell suface DPPIV and sucraseisomaltase were covalently modified by iodinated sulfo-SHPP which was recently introduced as an ideal reagent for the labeling of cell surface proteins (Thompson et al., 1987). The experiment was done analogous to that shown in Fig. 7. After radioiodination the cells were incubated at 37 "C for the indicated times and then lysosomes were purified. The lysosomes were solubilized and sucrase-isomaltase and DPPIV were immunoprecipitated and analyzed by SDS-PAGE followed by autoradiography (Fig. 8). The results of this experiment confirm that the two hydrolases are transported to lysosomes at only a low rate. While low amounts of DPPIV appeared in lysosomes (A), no sucrase-isomaltase was detect-able in this fraction even after an overnight incubation at 37 "C (B). Quantification of these autoradiographs ( Fig. 9) showed that the upt.ake of cell surface iodinated DPPIV was comparable to that of ""I-Fab fragments bound at the cell surface. After an overnight incubation the amount of DPPIV delivered to lysosomes was slightly higher when determined by the direct iodination procedure than by the Fab method. This might indicate that in lysosomes the antibodies have a shorter half-life than DPPIV. On the other hand, the experiment with sucrase-isomaltase suggests that the Fab assay is somewhat more sensitive than the iodination assay. Another important result emerging from Fig. 8 is that without an incubation at 37 "C neither DPPIV nor sucraseisomaltase were detected in the lysosomal fraction LII. Thus, the [""Slmethionine-labeled molecules found in the LII fraction after metabolic labeling cannot be due t.o cont.aminat.ion with brush border membranes.
Assuming that newly arriving enzyme molecules in the brush border are indistinguishable from those that already reside in this membrane these results suggest that endocytosis cannot fully explain the relatively high amount of brush border enzymes in lysosomes (see "Discussion").
Recycling of Hydrolases to the Cell Surface via Endosomes-The low rate of endocytic delivery of sucrase-isomaltase and DPPIV to lysosomes (Figs. 7-9) contrasts with the total rate of endocytosis (Fig. 6) and supports the notion that the two hydrolases recycle to the cell surface via endosomes. In order to test if brush border enzymes are internalized into endosomes, we applied Percoll gradient centrifugation to separate endosomes from lysosomes (Marsh et al., 1987). Our separation procedure was similar to that for lysosomes with the exception that a higher Percoll concentration was used (see "Materials and Methods" for details). Fig. 10 demonstrates that the separation of endosomes (labeled by allowing fluidphase endocytosis of horseradish peroxidase for 15 min) from lysosomes (detected by measuring glucosaminidase activity) was efficient in Caco-2 cells. Lysosomes were found near the bottom of the gradient while endosomes almost cofractionated with the brush border membrane marker alkaline phosphatase at the top of the gradient.
This fractionation procedure was applied to cells after lZ51-Fab fragments as already described in the experiments illustrated in Fig. 6. After centrifugation the Percoll gradients were fractionated and the fractions were pooled as indicated in Fig. 10. The time-dependent appearance of radioactivity in pools I (endosomal fraction) and III (lysosomal fraction) is given in Fig. 11. The results demonstrate that most of the endocytosed radioactivity entered endosomes but not lysosomes. The shape of the curves for pool I is similar to that found for total endocytosis, with the exception that both hydrolases reached a plateau in the endosomal fraction. However, when for each time point the radioactivity of pools I and III were added (correcting for the different yields of markers in the two fractions which were twice as large in pool I than pool III) curves were obtained that displayed identical shapes as those in Fig. 6 (not shown). These data are in line with the assumption that the majority of endocytosed sucraseisomaltase and DPPIV recycles to the cell surface rather than being transported to lysosomes. In order to test this presumed reappearance at the cell surface directly, we combined the Fab assay with NHS-SSbiotin, a reagent which allows reversible biotinylation of proteins. The experiment was started by labeling the cells at 4 "C with reversibly biotinylated Fab fragments specific for DPPIV. This experiment was performed with DPPIV only since the amount of internalized sucrase-isomaltase is too low to be analyzed. After incubating the cells at 37 "C for 75 min, the biotin of Fab fragments that had remained at the cell surface was cleaved off by reduction with glutathione and the cells were returned to the 37 "C! incubator.
After different times the cells were cooled to 4 "C and labeled with lz51streptavidin. Thereafter the cells were harvested and the cellassociated radioactivity was measured (Fig. 12). The results show that there is a time-dependent reappearance of biotinylated Fab fragments at the cell surface which levels off after 1 h. This flattening of the curve was to be expected since the internal pool of biotinylated Fab fragments becomes depleted after a certain time and since the Fab fragments reappearing at the cell surface are endocytosed again. It is important to note that 'Y-streptavidin is not limiting in this assay. Caco-2 cells were labeled and subjected to low pH treatment as described in Fig.  8. After homogenization the cells were fractionated on Percoll gradients and the fractions were pooled as indicated in Fig. 10. The controls were labeled with "'1-Fab fragments at 4 "C. The cells were harvested and total cell-associated radioactivity was determined.
Values are given as percent of control without correcting for the yields of marker enzymes. (x,pool I;0,pool III). DISCUSSION This study demonstrates that newly synthesized microvillar hydrolases are transported to lysosomes along two different pathways. A first fraction is delivered directly to lysosomes bypassing the brush border membrane, whereas some enzyme molecules endocytosed from the apical membrane appear in lysosomes in a second later wave. The majority of endocytosed sucrase-isomaltase and DPPIV was found to recycle to the brush border membrane.
The purity of the lysosomal fraction LII is critical for the validity of the conclusions drawn in this study. For instance, the appearance of metabolically labeled sucrase-isomaltase and DPPIV in LII may simply be due to brush border vesicles or Golgi-derived membranes cofractionating with lysosomes. Four observations argue against such a problem. First the enzymatic properties of the LII fraction suggest little crosscontamination by organelles of non-lysosomal origin. Second, transport rates (i.e. half-maximal appearance) of these enzymes to the Golgi apparatus and to the brush border membrane are different from those to the lysosomal fraction. In fact, the rate of transport of the two hydrolases to the brush border (Stieger et al., 1988) was at least 2 h faster in comparison to that to lysosomes. Third, when Caco-2 cells were surface-labeled at 4 "C with lZ51-Fab fragments specific for x t/ FIG. 12. Reappearance of internalized DPPIV at the cell surface.
Cells were labeled at 4 "C with biotinylated Fab fragments against DPPIV followed by a 75-min incubation at 37 "C. Then the cells were cooled to 4 "C! and the biotin of the Fab fragments that had remained at the cell surface was cleaved by reduction. The stripped cells were returned to the 37 "C incubator. After different time intervals the cells were cooled again and labeled with '251-streptavidin. After harvesting the cell-associated radioactivity was measured.
either DPPIV or sucrase-isomaltase and were fractionated without a chase the radioactivity associated with the lysoso-ma1 fraction was close to the detection limit of the y-counter. Fourth, it was not possible to immunoprecipitate cell surface iodinated proteins from the lysosomal fraction without a chase. The results obtained by the cell surface labeling approach also argue against a cross-contamination by endosomes. At steady state, 18% of the initially labeled DPPIV was found in kinetically early endosomes whereas this fraction was only 2% for sucrase-isomaltase. In contrast, the percentage of newly synthesized sucrase-isomaltase and DPPIV in lysosomes was comparable. These results rule out the possibility of a significant contamination of the lysosomal fraction by early endosomes. We are confident therefore that our lysosomal fraction is sufficiently pure to draw valid conclusions on the trafficking of brush border enzymes.
Lysosomal Delivery of Brush Border Hydrolases-Since metabolically labeled sucrase-isomaltase and DPPIV appeared in lysosomes later than in the microvillar membrane it was initially not possible to decide whether they are transported to lysosomes via the brush border membrane or by a direct intracellular pathway. The biphasic profile of appearance in the LII fraction, however, already indicated that the two hydrolases may be transported to lysosomes along two different pathways. Therefore, we have also studied the endocytosis route by using lz51-Fab fragments and covalent iodination to label cell surface sucrase-isomaltase and DPPIV in the microvillar membrane of intact cells. The Fab cell surface labeling procedure is highly selective as assessed by a number of control experiments that have been discussed under "Results." The kinetics of transport of cell surface sucrase-isomaltase and DPPIV to lysosomes was found to be different from that of metabolically labeled enzymes. Indeed, endocytic transport to lysosomes is too low and too slow to account for all of the metabolically labeled sucrase-isomaltase and DPPIV measured in this organelle. However, the endocytic kinetics were roughly concordant with the second wave of 35S-labeled hydrolases in lysosomes (that is at chase times longer than 10 h). This suggests that the first wave of [35S] met-labeled hydrolases in lysosomes was imported along a route that bypasses the brush border membrane. The only low level of delivery to lysosomes of cell surface-bound Fab fragments is unlikely to be due to dissociation of the immunocomplexes in acidic endocytic organelles since more than 80% of Y-Fab remained associated with the corresponding antigen after prolonged exposure to pH 3.0 (not shown). It is known that lysosomes, the most acidic organelles of the endocytic pathway, have a pH not lower than 4.5 (Mellman et al., 1986). Furthermore, the same degree of lysosomal delivery was obtained by iodinating cell surface DPPIV covalently. These biochemical findings were confirmed by immunoelectron microscopy using ultrathin cryosectioning. It was found that surface bound anti-sucrase-isomaltase or anti-DPPIV antibodies were internalized into endosomal structures while transport to lysosomes was s10w.~ These experiments also showed that antibodies to DPPIV or sucraseisomaltase label the microvillar membrane of intact Caco-2 cells in a uniform manner.
Recently, we have demonstrated that newly synthesized proteins are transported to the brush border membrane of Caco-2 cells along two different pathways: a direct intracellular and an indirect pathway via the basolateral membrane.' Therefore, the first wave of newly synthesized microvillar enzymes in lysosomes may be due to endocytosis from the basolateral membrane rather than a direct intracellular pathway. We have approached this possibility by a combination of the here described subcellular fractionation procedure and a selective cell surface biotinylation assay.* Unfortunately, no conclusive result was obtained due to extremly low radioactivity signals which were close to the detection limit. However, the following observations argue against the possibility that the lysosomal appearance is mainly due to a delivery of newly synthesized DPPIV and sucrase-isomaltase from the basolat-era1 membrane. 1) Whereas much more DPPIV than sucraseisomaltase was found in the basolateral membrane, similar amounts of the two enzymes were found in lysosomes. 2) DPPIV inserted into the basolateral membrane was shown to be transcytosed to the brush border membrane quantitatively. 3) Appearance in lysosomes occurs before disappearance in the basolateral membrane. Of course, a minor contribution from the basolateral membrane cannot be excluded.
Since the apparent molecular mass of the hydrolases in lysosomes is indistinguishable from that of the complexglycosylated forms in the homogenate it can be concluded that the hydrolases traverse all subcompartments of the Golgi apparatus before they are sorted to lysosomes. Therefore, they may migrate along the same route as resident lysosomal proteins. Lysosomal enzymes (Kornfeld, 1987;Griffiths and Simons, 1986) and lysosomal membrane glycoproteins (Green et al., 1987) are sorted from plasma membrane proteins at a tram+ or post-Golgi intracellular site, most likely in the trans-Golgi network and are transported to lysosomes via a specialized endosomal structure (prelysosome) (Griffiths et al., 1988). It is interesting to note that the transport of lysosomal membrane glycoproteins in kidney cells, macrophages, and Caco-2 cells is considerably faster than the delivery of sucraseisomaltase and DPPIV to lysosomes (Green et al., 1987).4 Recently, a striking homology was found between lysosomal a-glucosidase, a soluble enzyme, and sucrase-isomaltase (Hoefsloot et al., 1988). In contrast to a-glucosidase, which carries the lysosomal targeting signal phosphomannose, the sucrase-isomaltase (like DPPIV) is not phosphorylated in Caco-2 cells4 and thus a missorting of sucrase-isomaltase to lysosomes due to sequence homology with a lysosomal enzyme appears unlikely. Furthermore, it is important to note that our mAbs to sucrase-isomaltase do not cross-react with lysosomal a-glucosidase. Overall, these kinetic and structural considerations suggest that the mechanism of transport of sucrase-isomaltase and DPPIV to lysosomes differ from that for soluble or membrane-bound lysosomal resident proteins. What may be the function of a direct intracellular pathway to lysosomes? It has been postulated that the lysosomes may have a regulatory function in the cell surface expression of brush border membrane proteins (Blok et al., 1984) similar to the crinophagic pathway for secretory vesicles in the pituitary of lactating rats (Smith and Farquhar, 1966). Another potential function of this pathway may relate to a late "productcontrol" mechanism which would remove proteins from the biosynthetic pathway that, for example, are not correctly folded and therefore have not acquired their normal enzymatic activity or have no proper signal for their transport to the microvillar membrane. A quality control system has already been described for the endoplasmic reticulum (Doms et al., 1988;Lippincott-Schwartz et al., 1988). At present it is unknown, however, if the brush border enzymes in lysosomes are enzymatically active. Interestingly, a fraction of newly synthesized complex-glycosylated sucrase-isomaltase and DPPIV has a long apparent residence time in a Golgi fraction of Caco-2 cells (Stieger et al., 1988). A slowly turning-over pool of hydrolases in the Golgi apparatus would be in line with both a regulatory or a quality control mechanism. Endocytosis and Recycling of Brush Border Hydrolases-A remarkable finding is that sucrase-isomaltase and DPPIV undergo interiorization by endocytosis. However, in Caco-2 cells most of the endocytosed sucrase-isomaltase and DPPIV are not delivered all the way to lysosomes but recycle to the cell surface without being degraded as demonstrated for DPPIV. The magnitude of endocytosis is about lo-fold higher for DPPIV than for sucrase-isomaltase. The endocytosis is not due to a cross-linking effect of divalent antibodies as reported by Louvard (1980) for aminopeptidase N in MDCK cells, since we used monovalent Fab fragments. The reason for the differential endocytic behavior is not known but it is possible that sucrase-isomaltase is less mobile due to interactions with cytoskeletal components similar to Na'/K+-ATPase in the basolateral membrane of MDCK cells (Nelson et al., 1987). The steady state level of internalized DPPIV suggests the existence of an internal pool which is most likely localized in endosomes. Molecules localized in this pool or in the brush border membrane are continuously exchanged. There is no obvious function for such an internal pool nor for the recycling of brush border hydrolases.
We have recently demonstrated that newly synthesized brush border hydrolases transiently inserted into the basolat-era1 membrane are transcytosed efficiently to the apical membrane whereas apical-to-basolateral transcytosis did not occur.' Together with the results of the present study these data suggest efficient sorting of endocytosed plasma membrane proteins in Caco-2 cells. Interestingly, when the brush border membrane of Caco-2 cells was labeled with a mAb against DPPIV, incubated at 37 "C, and then processed for electron microscopic immunocytochemistry the endocytosed antibody was never found in the Golgi apparatus but in various endosomal structures3 This finding indicates that sorting of endocytosed plasma membrane proteins does not occur in the trans-Golgi network and supports the hypothesis that endosomes are of high importance for the biogenesis and maintenance of epithelial cell polarity.