Acidification and Ion Permeabilities of Highly Purified Rat Liver Endosomes*

While it is well established that acidic pH in endosomes plays a critical role in mediating the orderly traffic of receptors and ligands during endocytosis, little is known about the bioenergetics or regulation of endosome acidification. Using highly enriched frac- tions of rat liver endosomes prepared by free flow electrophoresis and sucrose density gradient centrifugation, we have analyzed the mechanism of ATP-de-pendent acidification and ion permeability properties of the endosomal membrane. This procedure permitted the isolation of endosome fractions which were up to 200-fold enriched as indicated by the increased specific activity of ATP-dependent proton transport. Acidification was monitored using hepatocyte and total liver endosomes selectively labeled with pH-sensitive markers of receptor-mediated endocytosis (fluorescein iso- thiocyanate asialoorosomucoid) or fluid-phase endocytosis (fluorescein isothiocyanate-dextran). In addi- tion, changes in membrane potential accompanying ATP-dependent acidification were directly measured using the voltage-sensitive fluorescent dye Di-S-C3(5). Our results indicate that ATP-dependent acidification of liver endosomes is electrogenic, with proton transport being accompanied by the generation of an inte-rior-positive membrane potential opposing further acidification. The membrane potential can be dissipated by the influx of permeant external anions or efflux of internal alkali cations. Replacement exter-nally of permeable anions with less permeable anions (e.g. replacing C1- with gluconate)

While it is well established that acidic pH in endosomes plays a critical role in mediating the orderly traffic of receptors and ligands during endocytosis, little is known about the bioenergetics or regulation of endosome acidification. Using highly enriched fractions of rat liver endosomes prepared by free flow electrophoresis and sucrose density gradient centrifugation, we have analyzed the mechanism of ATP-dependent acidification and ion permeability properties of the endosomal membrane. This procedure permitted the isolation of endosome fractions which were up to 200-fold enriched as indicated by the increased specific activity of ATP-dependent proton transport. Acidification was monitored using hepatocyte and total liver endosomes selectively labeled with pH-sensitive markers of receptor-mediated endocytosis (fluorescein isothiocyanate asialoorosomucoid) or fluid-phase endocytosis (fluorescein isothiocyanate-dextran). In addition, changes in membrane potential accompanying ATP-dependent acidification were directly measured using the voltage-sensitive fluorescent dye Di-S-C3(5).
Our results indicate that ATP-dependent acidification of liver endosomes is electrogenic, with proton transport being accompanied by the generation of an interior-positive membrane potential opposing further acidification. The membrane potential can be dissipated by the influx of permeant external anions or efflux of internal alkali cations. Replacement externally of permeable anions with less permeable anions (e.g. replacing C1with gluconate) diminished acidification, as did replacement internally of a more permeant cation K+ with less permeant species (such as Na' or tetramethylammonium). ATP-dependent H+ transport was not coupled to any specific anion or cation, however. The endosomal membrane was found to be extremely permeable to protons, with protons able to leak out almost as fast as they are pumped in. Thus, the internal pH of endosomes is likely to reflect a dynamic equilibrium of protons regulated by the intrinsic ion permeabilities of the endosomal membrane, in addition to the activity of an ATP-driven proton pump.
are internalized via coated pits and coated vesicles and then rapidly delivered to endosomes where ligands typically dissociate from their receptors. Free receptors are then recycled back to the plasma membrane while the discharged ligands are delivered to lysosomes and degraded (1,2). This molecular sorting of receptors and ligands is facilitated by the acidic pH found in endosomes, an environment which favors the disruption of many receptor-ligand complexes (2, 3). Dissipation of the pH gradient, using ionophores or acidophilic weak bases, generally inhibits receptor-ligand dissociation and rapid receptor recycling (3).
Like lysosomes and many secretory organelles, endosomes lower their internal pH via an NE"-sensitive proton ATPase (3)(4)(5). However, the internal pH in these various organelles varies considerably (from pH 4.6 to >6.5) and is even thought to vary between different endosome subpopulations (3,6). A number of experiments performed using intact cells have shown that internalized macromolecules encounter a progessively lower pH as they traverse the endocytic pathway en route to lysosomes (7)(8)(9)(10)(11). Thus, endosomes at different stages of maturation differ in their internal pH, indicating that endosome acidification is likely to be subject to regulation. Since endosomal pH may determine the rate and intracellular site at which receptor-ligand dissociation occurs, the regulation of endosome acidification is likely to play a critical role in controlling the intracellular traffic of internalized receptors and ligands (2,3).
In spite of considerable interest, few details are known concerning the actual mechanism by which endosomes lower or control their internal pH. This has been largely due to difficulties in preparing sufficient quantities of highly purified, functionally intact organelles to permit the detailed study of the proton transport and ion permeabilities of the endosomal membrane. Recently, we found that free flow electrophoresis (FFE) provides a rapid and effective method for the isolation of active endosome populations from tissue culture cells (12,13). In this paper, we show that FFE can be used in conjunction with sucrose density gradient centrifugation for the isolation of endosomes from rat liver in sufficient quantity to permit a more complete understanding of how endosomes lower their internal pH. Our results demonstrate that ATPdependent acidification in endosomes is electrogenic and is controlled not only by the activity of the H'-ATPase itself, but also by the ion permeability characteristics of the endo-soma1 membrane.

MATERIALS AND METHODS
Animals-Male Sprague-Dawley rats, 180-200 g, were purchased from the Charles River Breeding Laboratories (Wilmington, MA) and were fasted 24 h before use.
Selective Labeling of Rat Liver Endosomes-Endosomes were labeled with FITC-conjugated endocytic tracers by in situ perfusion of rat liver (17). Livers were equilibrated with perfusion medium (Hank's B) at 37 "C for 5 min, followed by labeling with FITC-dextran (5 mg/ ml in Hank's B, perfusate flow at 30 ml/min) for another 5 min. To remove extracellular marker and to inhibit further endocytosis, icecold buffer was then immediately circulated through the liver for an additional 3 min. Alternatively, endosomes were labeled by perfusion with FITC-ASOR (5 pg/ml in Hank's B) for 10 min, in the presence or absence of a 10-fold excess of non-derivatized ASOR to determine specific versus nonspecific uptake of the fluorescent probe. Release of remaining surface-bound ligand was achieved by addition of 10 mM Na+-EDTA to the ice-cold perfusion medium.
Preparation of Endosomes by Sucrose Density Gradient Centrifugation and Free Flow Electrophoresis-All manipulations were carried out at 0-4 "C unless otherwise indicated. Following endosome labeling, the liver (approximately 10 g) was quickly removed, minced, and homogenized in 30 ml of TEA-sucrose (0.25 M sucrose, 10 mM triethanolamine, 1 mM EDTA, 10 mM acetic acid, titrated with NaOH to pH 7.4 (12); for initial homogenization, buffer also contained a mixture of protease inhibitors including aprotinin, leupeptin, antipain, and chymostatin) using 8 strokes of a loose fitting Dounce homogenizer. A postnuclear supernatant was prepared by sequential centrifugation at 1,300 and 3,200 X g for 10 min each. The pellets (P1 and P2) were discarded and the final supernatant spun at 30,000 x g. , for 10 min yielding pellet (P3) and supernatant (S3) fractions. S3 was diluted with TEA buffer to the initial homogenate volume and centrifuged 1 h at 100,000 X g. , to prepare a crude microsomal pellet (P4). This pellet (P4), or alternatively P3, were used to prepare an enriched Golgi fraction, which was found also to be enriched in endosomes, by flotation in discontinuous sucrose density gradients (18): P3 or P4 were resuspended by 10 strokes in a tight fitting Dounce homogenizer in TEA buffer, and 2.0 M sucrose was added to a final concentration of 1.15 M. 15 ml of the resuspended pellets were placed in 38-ml centrifuge tubes, overlayed with 7 ml each of 1.0, 0.86, and 0.25 M sucrose, and centrifuged in a SW 28 rotor (Beckman) at 80,000 X g. , for 200 min. Material at the 0.25/0.86 interface (L1,Gl) and at the 0.86/1.0 interface (L2,G2) were analyzed for ATP-dependent acidification. As described previously (18), "L" fractions were derived from P3 (i.e. the 30,000 X g pellet) and "G" fractions were derived from the microsomal pellet (P4).
For separation of endosomes by free flow electrophoresis (FFE), the G1 fraction was concentrated by centrifugation (100,000 X g, 1 h) and resuspended in TEA buffer to a final concentration of 1 mg of protein/ml (19). As described in detail for the isolation of endosomes from tissue culture cells (12, 13), the sample was next subjected to gentle trypsin treatment by incubation for 5 min at 37 'C with 0.2% L-1-tosylamido-2-phenylethyl chloromethyl ketone-trypsin (mg of trypsin/mg of protein). The reaction was stopped by adding a 10-fold excess of soybean trypsin inhibitor at 0 "C and the sample was then injected into a FFE chamber at 2 ml/h using a chamber buffer-flow rate of 3 ml/fraction/h (12). FFE was performed using a Bender and Hobein ElphorVap 21 (Bender & Hobein, Munich, Federal Republic of Germany) at 130 mA and 1800 V using TEA-sucrose as chamber buffer (12, 13).
Cell-free Acidification Assay-ATP-dependent acidification of isolated endosomes labeled with pH-sensitive fluorescent probes (FITCdextran, FITC-ASOR) was carried out as previously described (4). All experiments were carried out using both the G1 sucrose gradient fraction as well as the more highly enriched endosome fraction obtained after free flow electrophoresis to determine if the electrophoresis procedure in any way altered the properties of the labeled endosomes. To calibrate the pH dependence of the fluorescence signal, FITC-ASOR-labeled endosomes were equilibrated with 150 mM KC1 buffer at various pH values in the presence of 10 p M nigericin to dissipate any preexisting pH gradients. To determine ATP-dependent acidification, endosomes (0.2-10 pg of protein) were equilibrated overnight on ice with 20 mM HEPES-tetramethylammonium hydroxide (TMA), pH 7.4,5 mM MgSO,, and 150 mM Na+ and/or K' salts. For ion substitution experiments, isomolarity was maintained by addition of sucrose, pH-dependent changes in FITC fluorescence were determined at ambient temperature using an LS-5 fluorometer (Perkin-Elmer) with excitation and emission wavelengths set at 485 and 515 nm, respectively. Acidification was initiated by addition of 2.5 mM ATP from a 500 mM stock of ATP adjusted to pH 7.4 with NaOH or KOH, as indicated. pH gradients were dissipated by adding 1 p~ of various ionophores (valinomycin, FCCP, nigericin) from 500 p~ stock solutions in ethanol. Addition of equivalent volumes of ethanol alone had no effect. In some experiments, ATP was removed by the addition of 10 mM glucose and hexokinase (12 units). Endosomes were equilibrated in the indicated buffers for 1 h or overnight at 4 "C or for 1 h at room temperature before ATP addition with equivalent results.
Membrane Potential Measurements-Alterations in the transmembrane potential were determined by following changes in the fluorescence intensity of the voltage-sensitive dye Di-S-Ca(5) at excitation and emission wavelengths of 620 and 670 nm, respectively (21). This positively charged fluorescent dye partitions between intra-and extravesicular space in response to a membrane potential. An intravesicular negative potential leads to dye uptake and to a decrease in fluorescence, whereas an inside-positive potential results in dye extrusion and hence to an increase in fluorescence. FFE-purified endosomes (40 pg) were pre-equilibrated overnight on ice with 100 mM Na-gluconate, 100 mM sucrose, 20 mM HEPES-TMA, pH 7.4, 5 mM MgS04. Di-S-C3(5) (0.25 pM) was added to a cuvette containing the pre-equilibrated endosomes. After stabilization of the Di-S-C3(5) fluorescence, 2.5 mM ATP (or another nucleotide triphosphate) was added. The resulting membrane potential was then dissipated with 1 p M FCCP. For inhibition experiments, 10 p~ NEM was included in the assay medium. Dye partitioning was calibrated by diluting preequilibrated endosomes into K+ buffers of higher or lower K+ concentrations in the presence or absence of valinomycin in order to establish diffusion potentials.

RESULTS
Selective Labeling of Rat Liver Endosomes-Prior to isolation, endosomes were selectively labeled by in situ perfusion of rat liver using well characterized markers of fluid phase or receptor-mediated endocytosis (4,14,22,23). For a pHsensitive fluid phase marker, FITC-dextran was perfused a t a concentration of 5 mg/ml (in Hank's B) for 5 min at 37 "C.
Since FITC-dextran would be expected to label endosomes in both parenchymal and non-parenchymal cells, we also used FITC-conjugated ASOR, which is internalized only by hepatocytes via the asialoglycoprotein receptor (22). For these experiments, endosome labeling was carried out by perfusion with 5 pg/ml FITC-ASOR for 10 min at 37 "C, conditions shown to result in the delivery of the ligand to endosomes and not to lysosomes (22, 23) (see below). Specific binding and uptake of FITC-ASOR was demonstrated by competition with a 10-fold excess of non-derivatized ASOR which reduced FITC-ASOR uptake to an amount undetectable in liver homogenates by fluorescence spectrophotometry (now shown).
Purification of Endosomes by FFE-Although selective labeling with pH-sensitive probes ensured that only endocytic vesicles would contribute to acidification activity observed in vitro (4), detailed analysis of the bioenergetics of ATP-driven proton transport still requires highly purified preparations of organelles. Because endosomes have a buoyant density similar to that of many other smooth membranes, separation of endosomes from Golgi, endoplasmic reticulum (ER), and plasma membranes has proved difficult using conventional techniques. Since FFE provides a rapid and effective method for the purification of endosomes from tissue culture cells (12), we next determined whether this approach could also be used to prepare much larger quantities of highly enriched endosomes from rat liver. A partially enriched endosome fraction was first prepared by differential and isopycnic centrifugation (18) prior to the FFE step. Using ATP-dependent acidification as a functional enzymatic marker for endosomes to follow purification, both FITC-dextran and FITC-ASOR-labeled endosomes were enriched in several previously described Golgi fractions (18). Briefly, after perfusion a postnuclear supernatant was prepared from the labeled liver homogenates and centrifuged to yield low speed 30,000 X g (P3) and high speed 100,000 X g (P4) pellets. These pellets were then fractionated by flotation in discontinuous sucrose density gradients (18). As shown in Table I, ATP-dependent acidification was most enriched at the 0.25/0.86 M ("Gl") and 0.86/1.0 M ("G2") sucrose interfaces. The G1 fraction exhibited the highest specific activity for ATP-driven proton transport, approximately 100-fold enriched relative to protein as compared to the starting material ( i e . the P4 pellet). With the exception of the trans-Golgi marker UDP-galactosyltransferase (40-fold enrichment), only slight enrichment of marker enzymes for organelles other than endosomes was found in the G1 fraction; therefore, G1 was selected for further purification by FFE. Importantly, lysosomes (@-hexosaminidase activity) were actually depleted in this fraction.
As previously found for preparation of endosomes from tissue culture cells (12, 13), optimal separation of rat liver endosomes by FFE required a brief treatment of the sample with low concentrations of trypsin. The endosome-enriched Golgi fraction (GI) was resuspended at 1 mg/ml protein, incubated with 0.2% trypsin/mg protein for 5 min at 37 "C followed immediately by the addition of a 10-fold excess of soybean trypsin inhibitor at 0 "C. The trypsin treatment affected neither acidification nor the protein profiles (as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis), but was absolutely required to achieve anodal deflection of endosomes away from the bulk of membrane and cytosolic proteins (12, 13,24).
The separation profile of G1 after FFE is shown in Fig. 1. FITC-ASOR-labeled endosomes migrated as a single peak which was well separated from a major unshifted peak containing UDP-galactosyltransferase activity (Golgi) and NADPH-cytochrome c reductase activity (ER). The specific activity of ATP-dependent acidification in the peak fraction (#35) was increased 2-3-fold relative to that measured in the starting G1 fraction (Table I). Marker enzymes for lysosomes (@-hexosaminidase) and mitochondria (not shown) were barely detectable after FFE. While a portion of the plasma membrane marker enzymes 5"nucleotidase (not shown) and alkaline phosphodiesterase I was associated with the unshifted peak, more than half of the alkaline phosphodiesterase I co-migrated at least in part with the anodally deflected endosomal fractions. However, alkaline phosphodiesterase I specific activity was 7-10-fold lower than for ATP-dependent acidification but represented <0.1% of the initial enzyme  Due to the turbidity of the initial homogenate, it was impossible to obtain an initial value for ATP-dependent acidification in this fraction (H). Therefore, ATP-dependent acidification is expressed as percent ATP-dependent FITC fluorescence quenching per mg of protein in each sample. Although arbitrary, comparison of these values provides a direct indication of relative enrichment of ATP-dependent acidification activity and, thus, of endosome purification. For example, the FITC-dextran containing endosomes in the G1 sucrose interface fraction, measured as having a specific activity of acidification of 100, were 100-fold enriched relative to protein compared to the P4 pellet fraction from which it was derived (specific activity -1.0). The FEE step resulted in an additional 2.7-fold enrichment. 100% of the acidification activity measured in P4 (and P3) could be recovered from the sucrose gradients; >70% of this activity was divided between the G1 (L1) and G2 (L2) fractions, with G1 ( L l ) having the higher specific activity (and containing 2-3-fold more total activity) as indicated in the Table. ND, not detected. 5'-Nucleotidase activity was not detected in fraction 35 due to insensitivity of the assay (relative to alkaline phosphodiesterase) and due to the severalfold dilution that occurs during the electrophoresis step. activity found in the homogenate or P4 fraction. 5"Nucleotidase activity, measured using a less sensitive assay, was not reproducibly detectable in the shifted fractions after FFE. The recovery of all enzyme activities, fluorescence, and protein from the G1 fraction after the FFE step was >90%.
Characterization of Rat Liver Endosomal Fractions-The degree to which the endosomes were enriched in the anodally deflected fractions was estimated by determining the recovery and relative specific activities of various organelle marker enzymes at each stage of purification (Table I). As mentioned above, the starting material, i.e. the G1 fraction from the sucrose flotation gradient, was already 90-100-fold enriched in ATP-dependent acidification (using either FITC-ASOR or FITC-dextran as endocytic markers) relative to the 100,000 X g pellet, significantly more than the enrichment of galactosyltransferase (40-fold), NADH-cytochrome c reductase (9fold), alkaline phosphodiesterase I (&fold), and 5"nucleotidase (%fold) or the lysosomal marker P-hexosaminidase (0.3fold). After separation by FFE, the endosomes were enriched an additional 2-3-fold in the anodally shifted fractions: final enrichment of 265 f 42-fold for FITC-ASOR-and 170 k 30fold for FITC-dextran-labeled endosomes. The apparent difference in enrichment for the two markers was not statistically significant but may reflect the fact that fluid-phase and receptor-bound endocytic tracers can accumulate in different endosome subpopulations.
More importantly, FFE further depleted the endosomal fraction of markers for ER, Golgi apparatus, plasma membrane (5' nucleotidase), lysosomes, and mitochondria (the latter were undetectable even in the G1 and G2 fractions prior to electrophoresis). Only alkaline phosphodiesterase I was enriched (25-fold) in the shifted endosome fraction, and presumably reflected that fraction of plasma membrane ectoenzymes which exist as intrinsic components of endocytic vesicle dities of Rat Liver Endosomes 2215 membrane in rat liver (25,26). This suggestion is supported by the fact that the enrichment of alkaline phosphodiesterase I in the G1 fraction by FFE was comparable (4-fold) to that observed for the endosome markers FITC-ASOR and FITCdextran. It is less likely that the co-migrating alkaline phosphodiesterase I represented contaminatingplasma membrane, since we have previously found that 9-labeled Chinese hamster ovary (CHO) cell surface proteins do not appear in the shifted endosome fraction after FFE, whereas alkaline phosphodiesterase I partially co-localizes with endosomes (12,13). In addition, the fraction of alkaline phosphodiesterase I in the final endosome peak was minor and represented <0.02% of the initial alkaline phosphodiesterase I activity in the homogenate; this contrasted with the approximately 20% recovery of FITC-ASOR in this fraction. 5'-Nucleotidase, determined using a less sensitive enzyme assay, was not reproducibly detected in the shifted fractions (which were diluted severalfold relative to GI). Isolated Endosomes Have an Acidic Internal pH-To determine whether the isolated endosomes retained an acidic internal pH in uitro, FFE-purified FITC-ASOR-labeled endosomes were next resuspended in an isosmotic buffer containing 150 mM KC1 and FITC fluorescence intensity monitored as a function of time. As shown in Fig. 2, the fluorescence signal increased gradually, reflecting a continual increase in internal pH. Using a pH calibration curve generated using intravesicular FITC-ASOR (not shown), the initial pH was estimated to be between 5.0 and 5.5 immediately after isolation. Dissipation of the initial pH gradient required about 60 min and was dependent on the presence of external cations (K' or Na+) as well as on their concentration (dissipation was significantly slower at concentrations <lo0 mM; not shown).
While the addition of the H+ ionophore FCCP to freshly isolated endosomes in the presence of KC1 only slightly enhanced the rate of proton efflux, the K+ ionophore valinomycin greatly stimulated H+ efflux, even in the absence of FCCP (Fig. 2). These data indicate that the endosomal mem- In each case, the initial FITC fluorescence intensity increased gradually, reflecting a dissipation of the initial acidic internal pH. While addition of the K+-ionophore valinomycin accelerated the decay of the pH gradient, the addition of the H' ionophore FCCP had little effect. After dissipation of the initial pH gradient, endosomes were re-acidified by addition of 2.5 mM K'-ATP (in the presence of MgC12). Presumably due to their increased conductance for H', the ATP-driven acidification of FCCP-treated endosomes was significantly reduced. A rapid increase in FITC fluorescence was obtained by adding the K+/H' ionophore nigericin (1 p~) which would be expected to immediately bring the internal endosomal pH to that of the suspending medium (ie. pH 7.4). brane has intrinsic permeabilities for H+ and K+ which allow electrically coupled exchange of internal H+ for external cations, thus allowing for the dissipation of the initial pH gradient in KC1 buffer. The fact that H' efflux was stimulated by valinomycin, but not by FCCP, indicated further that the K' permeability, and not the H' permeability, was ratelimiting for dissipation of the H+ gradient. Thus, the FITC-ASOR-containing endosomes were more permeable to H' than to K' . As expected, addition of the K+/H+ ionophore nigericin led to a rapid release of internal H+ (Fig. 2).
Importantly, identical results were obtained (for this and all subsequent experiments involving FITC-labeled probes) when the G1 and G2 fractions were used without further enrichment by FFE. This demonstrated that neither the trypsin treatment needed to prepare samples for FFE nor the electrophoresis itself had any effect on H+ or K+ permeability of FITC-ASOR or FITC-dextran-labeled endosomes from rat liver.
Characteristics of the Endosomal Proton Pump-Adding ATP to acidic endosomes resulted in no further decrease in FITC fluorescence, suggesting that there is a maximum pH gradient that can be established. Therefore, all experiments designed to characterize the ATP-dependent acidification of endosomes were performed under conditions of ionic equilibrium, i.e. after incubation of the endosomes for 1-24 h at 4 "C in cation-containing buffers to dissipate any pre-existing H+ gradients (Fig. 2). For all experiments, equivalent results were obtained as long as the internal pH was permitted to equilibrate with that of the suspending medium, irrespective of the length of preincubation. In fact, FFE-purified endosomes could be incubated at 4 "C for up to 80 h without loss of acidification activity. As shown in Fig. 2, addition of ATP to endosomes previously allowed to dissipate their pre-existing pH gradients exhibited a rapid nigericin-reversible decrease in FITC fluorescence. ATP-driven acidification was observed in FCCP-treated endosomes, indicating that even when freely permeable to H' , the endosome's limited K+ permeability was sufficient to stabilize a transmembrane pH gradient via an electrochemical potential.
We first examined the nucleotide specificity of the endo-soma1 proton pump. Acidification was found to be specific for ATP, as earlier observed for crude preparations of endosomes from tissue culture cells (10). Addition of other nucleotides such as CTP, GTP, or the nonhydrolyzable ATP analog AMP-PNP, did not support acidification of FFE-purified endosomes (not shown).
Next, we examined the effects of several potential H+-ATPase inhibitors on acidification. Total inhibition of ATPdriven proton transport was achieved after 5 min incubation with 10 p~ NEM, the best characterized inhibitor of vacuolar H' pumps (3, 5). Inhibitors of either the mitochondrial-type FIFo-ATPase (efrapeptin) or the gastric K+/H+-ATPase (Na3V04) had no effect on acidification, as described for endosomes from tissue culture cells (not shown) (4).
The Endosomal Proton Pump Is Electrogenic-NEM-sensitive proton pumps present in clathrin-coated vesicles, estrogen-induced rat liver multivesicular bodies, lysosomes, secretory granules, and Golgi membranes have been shown to operate by an electrogenic mechanism, H' transport being accompanied by an inward movement of positive charge (3, [27][28][29][30]. To determine whether the rat liver endosomal H' pump was also electrogenic, acidification activity was measured in FFE-purified endosomes equilibrated in buffers containing either KC1 or K-gluconate. Replacing C1- (Fig. 3a) with the less permeant organic anion gluconate (Fig. 3b), inhibited both the rate and extent of ATP-induced acidifica-  C1-( p a n e l a ) with the less permeant organic anion gluconate (panel b ) inhibited acidification. This inhibitory effect could be overcome by the K+ ionophore valinomycin. Direct measurement of potential changes in endosomes with a voltage-sensitive dye Di-S-C3(5) is shown in panels c and d. In panel c, K+-induced diffusion potentials were used to verify the ability of the dye to partition in accordance with predictable differences in transmembrane potential. FFE-purified endosomes (40 pg of protein) were equilibrated with 10 mM K+-gluconate overnight and then diluted into buffer containing 1 mM K+-gluconate, 10 mM K+-gluconate, or 100 mM K+-gluconate (as indicated) and 0.25 p~ Di-S-C3(5). 10 p M valinomycin was then added to each cuvette to facilitate the further development of a diffusion potential. As shown, the dye appropriately reports both positive and negative potential changes following dilution into buffers of higher and lower [K+], respectively. In panel d, endosomes were equilibrated with 100 mM Na+-gluconate and diluted into a cuvette containing the same buffer and 0.25 p~ Di-S-Ca(5). After the fluorescence signal reached a stable baseline, 1 mM ATP was added. The ATP-dependent increase in Di-S-C.45) fluorescence, indicating the formation of an interior-positive potential, was reversed by 1 p~ FCCP, demonstrating that the potential difference was due to H+. Addition of 10 p~ NEM inhibited the ATP-induced increase in Di-S-Ca(5) fluorescence. tion, suggesting that passive influx of external anions helps to dissipate an interior-positive membrane potential due to H+ influx. Consistent with this interpretation, addition of the K' ionophore valinomycin to gluconate-equilibrated endosomes (Fig. 3b), completely restored the acidification observed in the presence of C1-, presumably by facilitating the efflux of internal K+ down the electrical gradient. As expected, addition of valinomycin to KC1-equilibrated endosomes had Ion Permeabilities of Rat Liver Endosomes 2217 no stimulatory effect (Fig. 3a). These data strongly suggest that ATP-dependent acidification of endosomes is characterized by electrogenic H' influx, which can be influenced by the membrane's permeability to anion or cations. T o provide further evidence that ATP-coupled proton transport was electrogenic, we next used the positively chargedpotential sensitive dye Di-S-C3 (5). This dye partitions between intra-and extravesicular space in response to the membrane potential (30). It is important to note that unlike the pH-sensitive FITC-labeled endocytic probes, this externally added voltage probe is nonselective and will partition across any membrane present. As a result, the use of Di-S-C3(5) is absolutely dependent on the availability of a homogenous, highly purified population of organelles.
We first validated that Di-S-C3(5) could detect changes in membrane potential in FFE-purified endosomes by calibrating the fluorescence signal after creating K+-valinomycin diffusion potentials. This was accomplished by first loading endosomes with 10 mM K-gluconate and diluting them into buffers of higher or lower K' concentrations. Dilution into a medium containing 1 mM K-gluconate generated an interiornegative membrane potential, due to KC efflux down the concentration gradient, and was accompanied by a decrease in Di-S-C3(5) fluorescence due to quenching resulting from the dye influx and/or binding to the endosomes (Fig. 3c). On the other hand, dilution from 10 mM K' into a buffer with a higher K' concentration (100 mM) generated an interiorpositive potential and which led to an increase in fluorescence. As expected, these differences were accentuated by the addition of valinomycin which increased the endosomes' permeability to K+ (Fig. 3c).
Having demonstrated that Di-S-C3(5) can be used to report changes in membrane potential, we next determined whether ATP-dependent proton pumping is accompanied by membrane potential alterations. After equilibrating endosomes in Na+-gluconate buffer containing 0.25 FM Di-S-C3(5), addition of ATP resulted in a rapid increase in fluorescence, reflecting dye efflux concomitant with acidification (Fig. 3d). This effect could be reversed by the H' ionophore FCCP, demonstrating that the fluorescence increase was indeed due to H' uptake.
In contrast, CTP and AMP-PNP, two nucleotides which do not support acidification, failed to alter the Di-S-C3(5) fluorescence signal (not shown). Likewise, 10 PM NEM inhibited the ATP-dependent fluorescence increase (Fig. 3d). Thus, the fluorescence signal was not generated as an artifact of nucleotide addition (21). Taken together with the anion replacement experiments (Fig. 3, a and b ) , these data demonstrate the existence of an electrogenic H' pump in rat liver endosomes. No attempt was made to quantify changes in membrane potential using Di-S-C3(5), however, due to quantitative inaccuracies characteristic of this method and due to the fact that although highly enriched, FFE-purified endosomes may still contain contaminating membrane vesicles of indeterminate volume capable of generating a signal with Di-S-Cs (5).
Influence of Ion Substitution on ATP-dependent Acidification-Since the interior-positive membrane potential due to H+-transport would be dissipated by the influx of permeable external anions or efflux of internal alkali cations, the extent of acidification will in part be determined by the ion permeability characteristics of the endosomal membrane. T o determine the anion and cation permeabilities of rat liver endosomes, we next tested the effect of anion and cation substitutions on ATP-dependent acidification under conditions of ionic equilibrium.
Anion Permeability-Endosome-enriched fractions were preincubated with the Na' salts of various anions prior to determining the rate and extent of ATP-induced acidification in media of identical compositions. As shown in Fig. 4a, acidification was supported to different extents by different anions. The rate and extent of acidification decreased in the following order: C1-> gluconate = sulfate = isethionate > NOT > SCN-. Thus, substitution of less permeant anions such as gluconate, sulfate, or isethionate for a more permeable anion such as C1-decreased acidification. Interestingly, however, inhibition of acidification occurred in the presence of highly permeable anions such as SCN-and NOT at concentrations ranging from 12 to 150 mM.
The effect of NOT appeared to be due to a direct inhibition of acidification activity since NOT was also found to block H+ transport in the presence of the permeant anion C1-(not shown). In addition, the addition of NOi to acidic endosomes after the addition of ATP led to the dissipation of the pH gradient. Indeed, NOT has previously been found to be an inhibitor of acidification of other organelles in plant and mammalian cells (3,31).
The inhibitory effect of SCN-on ATP-dependent acidification, however, is difficult to reconcile with its normally high membrane permeability that might be expected to stimulate acidification. Indeed, such stimulation has been observed for ATP-driven H' transport in renal cortical endocytic vesicles and Golgi vesicles (21, 32, 33).
Cation Permeability-We next investigated the endosomal membrane's cation permeability by determining the effect of cation substitution on ATP-dependent H' transport. FITCdextran-containing endosome-enriched Golgi fractions were pre-equilibrated with the C1-salts of various cations for 24 h on ice prior to the determination of ATP-induced H' transport. As shown in Fig. 4b, ATP-driven H+ transport was stimulated by Na+ > K+ > choline > TMA'. These data demonstrate that ATP-dependent endosome acidification has no absolute or specific cation requirement, unlike the gastric K+/H+-ATPase, for example. The differential ability of the various cations to support electrogenic H' almost certainly  reflects the differential membrane permeabilities to these ions since continued H' influx will be compensated by the efflux of internal cations. Therefore, the more permeant the internal cation, the more acidification. Accordingly, the endosomal membrane is more permeant for Na' and K' than for choline and TMA'.
It is thus likely that both anion and cation permeabilities play a role in ATP-dependent acidification. Even in medium containing C1-, acidification is decreased in the presence of impermeant cations such as TMA, and vice versa, in the presence of Na'or K+-gluconate.
Ion Permeabilities Can Regulate ATP-driven Proton Transport-We have demonstrated that the activity of the endo-soma1 proton pump can be regulated by the development of an interior-positive membrane potential which, in turn, can be regulated by the intrinsic ion conductances of the endosomal membrane. Since the endosomal membrane also exhibits a significant conductance for protons (Fig. 2), it is possible that electrically coupled H'/K' (or Na'/H') exchange may also serve to control internal pH in endosomes. This was evaluated by observing the rate of dissipation of an ATPgenerated pH gradient in buffers containing either K+ or Na+.
FFE-purified endosomes were equilibrated in 25 mM KC1 or 25 mM NaCl and H' transport was initiated by adding ATP. When maximum acidification had been obtained, ATP was "removed" by the addition of glucose and hexokinase, which resulted in a rapid dissipation of the pH gradient (Fig.  5a). The half-time of decay of the pH gradient was -2.5 min in K' and -4 min in Na' .
Similar results were obtained when, instead of removing ATP, the endosomes were rendered freely permeable for H' by the addition of FCCP (Fig. 56). Under these conditions, where cation permeability was the only rate-limiting factor affecting H' efflux, the decay of the pH gradient was again faster in K' than in a Na' buffer. Taken  to K' than to Na'. Hence, the reduction of acidification activity in K+ buffer even under conditions of ionic equilibrium (Fig. 4b) can be explained by a significantly higher permeability to K' which favors H+ efflux by electrically coupled K+/H+ exchange. Fig. 5 also demonstrates that endosomes are significantly permeable to H' , indicating that acidification is likely to be associated with a dynamic flux of H' .
To directly demonstrate the effect of ion permeabilities on acidification, experiments were also performed with different ion compositions on both sides of the membrane. Endosomes were equilibrated with NaCl and diluted into a KC1 buffer or vice versa. As shown in Fig. 6a, when the more permeable cation (i.e. K' ) was inside, acidification was favored. Thus, optimizing conditions for the efflux of internal cations facilitates ATP-driven H' influx.
The effect of anion gradients on ATP-dependent acidification was also investigated. Endosomes were loaded with Na'gluconate or NaCl and diluted into NaCl or Na+-gluconate. As shown in Fig. 6b, when the more permeable anion was outside, acidification was favored. This result may also reflect a direct stimulation of the endosomal proton pump by external C1-, as demonstrated for the proton pump isolated from bovine adrenal chromaffin granules (34).
Finally, the general potentiating effect of permeable anions and cations on acidification is illustrated in Fig. 6c. completely omitted, i.e. using endosomes in isosmotic sucrose. When endosomes were equilibrated with Na+-gluconate or K+-gluconate and diluted into a medium containing either sucrose or TMA-C1, the presence of the more permeant cation inside, or the more permeant anion outside, had a potentiating effect on ATP-dependent acidification. Thus, it is clear that both cation and anion permeabilities can play important roles in regulating acidification by an electrogenic H+ pump.

DISCUSSION
The bioenergetics of endosome acidification was investigated using highly purified rat liver endosomes selectively labeled with pH-sensitive endocytic tracers. We have shown that the rat liver endosomal H+ pump operates by an electrogenic mechanism and thus belongs to the class of NEMsensitive vacuolar ATPases (3). Due to the high H+ permeability of the endosomal membrane, continuous inward H+ pumping was required to maintain the pH gradient. In addition, we found that the magnitude of the pH gradient established was influenced by the permeability of the endosomal membrane for cations and anions. Together, these findings demonstrate that acidification is associated with an unexpectedly dynamic flux of H' across the endosomal membrane. Thus, the net accumulation of intravesicular H+ at equilibrium must therefore reflect both the rate of ATP-driven H+ pumping and the membrane's other permeability characteristics. The basic features of rat liver endosome acidification are similar to those described for two other endocytic vesicle populations, estrogen-induced rat liver multivesicular bodies (28,31) and renal cortical cell endocytic vesicles (21), although the origin, purity, and function of these earlier preparations were less certain.
Our detailed investigation was facilitated by a new approach to endosome isolation, FFE, which has previously been applied only to tissue culture cells (12, 13). We have shown that FFE, in combination with sucrose density gradient centrifugation, can be successfully used for the isolation of a endosomal fraction from rat liver which is essentially devoid of other intracellular organelles known to contain H' pumps, such as Golgi, lysosomes, and mitochondria. Although the electrophoretic separation is dependent on a brief trypsin treatment, trypsinization had no quantitative or qualitative effect on ATP-dependent acidification. With the exception of experiments involving the voltage dye Di-S-CJ5) which required the use of FFE-purified organelles, identical acidification and permeability data were obtained using endosome-enriched G1 or G2 sucrose gradient fractions without trypsin treatment. The acidification activity of FFE-purified endosomes was, in fact, more stable than that found in the G1/G2 fraction, permitting extended preincubation times (up to 80 h) without loss of activity. Similar results were obtained for the acidification and permeability properties of endosomes isolated from CHO cells before and after trypsin treatment or FFE (13,24,35). Furthermore, detailed analysis of the effects of trypsin on the protein composition of endosomes isolated from CHO cells failed to detect any significant alterations (24). While trypsin treatment, FFE, or time of preincubation at 4 "C had any affect on the bioenergetics of isolated endosomes, we cannot exclude the possibility that some other aspect of an isolation procedure may produce somewhat different results in vitro from that found in intact cells.
The properties of the endosomal H+-ATPase are quite similar to the characteristics of vacuolar-type H+ pumps found in other organelles of the endocytic and biosynthetic pathways such as coated vesicles, lysosomes, and acidic secretory granules (3, 5 , 27-30). The rat liver endosomal H' pump is electrogenic and does not exhibit an absolute requirement for specific anions and cations, although H+ transport does not take place when ions are totally omitted. As demonstrated for the H+ pump of estrogen-induced multivesicular bodies, as well as for vacuolar H+-ATPases in plants (3,311, the rat liver endosomal H+ pump is blocked by nitrate, presumably by direct inhibition of the ATPase. Unexpectedly, however, the highly permeant anion SCN-was also observed to inhibit ATP-driven acidification. Therefore, we assume, that SCNmust exert a specific deleterious effect on rat liver endosomes, either by altering endosome permeability or by directly inhibiting the proton pump (as found for the chromaffin granule H+-ATPase (34)).
Mechanisms of pH Regulation in Endosomes-Internalized ligands en route to lysosomes are believed to encounter endosomes of progressively lower pH, ranging from pH > 6.3 to 5 (2, 6-11). Since different receptor-ligand complexes may dissociate at characteristic pH values, and since ligand dissociation can influence the pathway and efficiency of receptor recycling (2,3), the regulation of pH in endosome subpopulations may play an important role in controlling membrane traffic during endocytosis. Given that all organelles along the endocytic pathway are likely to contain the same H+ pump, other mechanisms must be provided to regulate the activity of this ATPase.
It seems unlikely that pH regulation is accomplished by controlling the number of ATPase molecules present in each compartment. Endosomes contain a small internal volume (approximately lo"* ml for a spherical endosome 0.5-pm in diameter); therefore only limited numbers of H+ are needed to lower the internal pH to <6. In addition, the observed pH differences among endosomes are relatively slight (<1-1.5 pH units). Since endosomes were found to be highly permeable to H+ in vitro and, in fact, leaked protons almost as fast as they were pumped in, continuous H+ pumping was required to maintain the pH gradient. If this situation approximates the permeability properties of endosomes in uiuo, a high rate of H+ flux would provide an ideal mechanism for controlling intravesicular pH in intact cells by regulating the net accumulation of internal H+ in dynamic equilibrium with cytosolic H+. The equilibrium would be subject to the permeability characteristics of the endosomal membrane since permeability to other anions and cations can control both the rate of H+ efflux by H+-cation exchange and the activity of the voltagesensitive, electrogenic H+-ATPase by alterations in membrane potential. It will now be important to determine whether the anion and cation permeabilities of the endosomal membrane are themselves subject to regulation (e.g. by phosphorylation) or whether different endosome subpopulations exhibit different permeability characteristics.
In addition, we recently found that endosomes which occur "early" on the endocytic pathway represent a distinct population involved in receptor recycling which is generally less acidic than kinetically "late" endosomes (13). Since early and late endosomes have markedly different protein compositions, it is indeed possible that these two subpopulations also have distinct ion permeability properties which are responsible for their different pH values. We have already identified at least one potential difference: the presence of the Na+,K+-ATPase only in early endosomes from CHO cells (34). The Na+,K+-ATPase appears to attenuate ATP-dependent acidification of early endosomes by facilitating the development of the interior-positive membrane potential which in turn inhibits the activity of the H+-ATPase. While such a situation may serve to help regulate pH in some cell types, we have not found similar evidence supporting a regulatory role for Na+-K+-Ion Permeabilities of Rat Liver Endosomes ATPase in rat liver endosomes.' The endocytic pathway in hepatocytes is complex, involving not only constitutive recycling of receptors at the sinusoidal (or basolateral) plasma membrane, but also the transport of certain receptors and ligands to lysosomes, the Golgi, or to the bile canalicular (apical) plasma membrane. The experiments discussed here have focussed only on the "total" endosome population labeled with fluid phase or receptor-bound markers destined primarily for lysosomes. While the basic bioenergetics of acidification of this population have now been defined, it will be important to define the properties of endosomes involved in the other transport pathways found in polarized epithelial cells.