Evidence for T w o Types of Rat Liver Microsomes with Differing Permeability to Glucose and Other Small Molecules*

Radioisotope flux measurements using Millipore fil- tration revealed two populations of rat liver microsomes designated type A and B. Type A and B vesicles are similar in that both are essentially impermeable to sucrose yet permeable to C1-. About 70% of the micro- somes (type A) are permeable to D-glUCOSe, L-glucose, 2-deoxy-~-glucose, D-mannose, D-mannitol, uridine, glycine, L-leucine, choline+, Tris+, Rb+, K’, and Na’. Other solutes such as D-gluconate-, D-glucosamine+, N-acetyl-D-glucosamine, L-glutamate-, L-lysine+, sul-fate2-, oxalate2-, and phosphate anions transverse type A vesicles with an intermediate rate. All of the above solutes except C1- pass with a comparatively slow rate the remaining 30% type B vesicles. Both type A and B microsomes are relatively impermeable to glucose 6- phosphate and related monophosphates. Membrane potential measurements using liver microsomes and con- trol membrane vesicles derived from rabbit skeletal muscle sarcoplasmic reticulum indicated that type A liver microsomes, despite being permeable to K+ and Na+, either lack or contain only a small number of highly conducting K+ and Na+ structures, such as the K,Na channel of sarcoplasmic reticulum. Treatment with the anion transport inhibitor 4,4’-diisothiocyanos-tilbene-2,2’-disulfonic acid

Radioisotope flux measurements using Millipore filtration revealed two populations of rat liver microsomes designated type A and B. Type A and B vesicles are similar in that both are essentially impermeable to sucrose yet permeable to C1-. About 70% of the microsomes (type A) are permeable to D-glUCOSe, L-glucose, 2-deoxy-~-glucose, D-mannose, D-mannitol, uridine, glycine, L-leucine, choline+, Tris+, Rb+, K', and Na'. Other solutes such as D-gluconate-, D-glucosamine+, N-acetyl-D-glucosamine, L-glutamate-, L-lysine+, sul-fate2-, oxalate2-, and phosphate anions transverse type A vesicles with an intermediate rate. All of the above solutes except C1-pass with a comparatively slow rate the remaining 30% type B vesicles. Both type A and B microsomes are relatively impermeable to glucose 6phosphate and related monophosphates. Membrane potential measurements using liver microsomes and control membrane vesicles derived from rabbit skeletal muscle sarcoplasmic reticulum indicated that type A liver microsomes, despite being permeable to K+ and Na+, either lack or contain only a small number of highly conducting K+ and Na+ structures, such as the K,Na channel of sarcoplasmic reticulum. Treatment with the anion transport inhibitor 4,4'-diisothiocyanostilbene-2,2'-disulfonic acid lowered the permeability of type A vesicles to several uncharged and negatively charged solutes including D-glUCOSe and gluconate-. These results suggest that a large fraction of liver microsomes is rendered permeable to various biologically relevant solutes and ions, perhaps through the presence of one or more channels with a maximal diameter of approximately 7-8 A which select(s) against solutes on the basis of their size and charge.
Several membrane permeation and transport systems are associated with the endoplasmic reticulum of liver cells. A calcium-sequestering system analogous to although less active than the well studied sarcoplasmic reticulum Ca'+-transport ATPase of striated muscle has been demonstrated in rat liver microsomes (Moore et al., 1975). Evidence has been presented that glucose-6-P hydrolysis is mediated by a two-component system consisting of a specific glucose-6-P permease which transfers glucose-6-P across the endoplasmic reticulum membrane and a nonspecific phosphohydrolase-phosphotransferase which is localized on the luminal side of the membrane (Ballas and Arion, 1977). Polypeptide and saccharide trans-* This work was supported by Research Grant AM18687 from the United States Public Health Service. 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. port proteins may be involved in the synthesis of structural and enzymatic membrane and/or extracellular proteins.
A NH2-terminal extension of about 20 relatively hydrophobic amino acids is believed to direct the growing polypeptide chain into the endoplasmic reticulum lumen (Blobel and Dobberstein, 1975), possibly via a polypeptide-permeable structure. In the case of N-glycosidically linked oligosaccharides, a dolichyl diphosphate lipid intermediate containing Nacetylglucosamine, glucose, and mannose has been implicated to serve as a direct precursor for the oligosaccharide chain of glycoproteins (Staneloni and Leloir, 1979). The lipid-saccharide intermediate is thought to be synthesized on the cytoplasmic side of the endoplasmic reticulum and after its movement across the membrane, to transfer the saccharide portion to the nascent polypeptide chain (Snider et al., 1980). Glucose and some of the mannose residues are subsequently released (Robbins et al., 1977;Tabas et al., 1978;Hunt et al., 1978). However, the mechanism(s) by which the lipid-saccharide intermediate and the released saccharides cross the membrane have not yet been resolved. Biosynthesis of 0-glycosidically linked oligosaccharides is also poorly understood. In particular, it is not clear whether the saccharides are attached to the protein before or after the nascent polypeptide chain has been transferred across the membrane. In either case, it may be speculated that the endoplasmic reticulum membrane contains structures which facilitate movement of hydrophilic saccharides across the membrane.
In the present study, we have investigated the nature of the permeability barrier formed by rat liver microsomes. We found that rat liver microsomes consist of two types of vesicles which differ in their permeability to various biologically relevant solutes. One group of vesicles (designated type A) is permeable to glucose and certain other small solutes, while the second type (type B) is relatively impermeable to these compounds.  and ~-[4,5-"H]leucine from New England Nuclear, Boston, MA. 4,4'-diisothiocyanostilbene-2,2'disulfonic acid' was purchased from Pierce, Rockford, IL. The fluorescent dye 3,3"dipentyl-2,2'-oxacarbocyanine was the generous gift of Dr. Alan S. Waggoner (Amherst College, Amherst, MA). Other reagents used were of reagent grade.

Reagents
Isolation ofMembranes"Microsoma1 membrane fractions derived from rat liver were prepared as follows. A rat fed a d libitum and weighing about 200 g was decapitated using a guillotine. The liver was rapidly excised, placed into ice-cold 0.25 M sucrose, and cut into small pieces. After transfer into 5 volumes of 0.25 M sucrose and 10 mM 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid (Hepes) (pH 7.51, the tissue pieces were homogenized at 0 "C using a smoothwalled Potter-type homogenizer with a Teflon pestle (A. H. Thomas & Co., Philadelphia, PA). The homogenate was centrifuged for 20 min at 10,000 rpm (7,800 X gav) in a Beckman type 35 rotor. A microsomal fraction was obtained from the supernatant by centrifugation for 60 min at 35,000 rpm (100,000 X gav) in a Beckman type 42 rotor. The upper, reddish part of the pellets was resuspended in 100 ml of 400 mosM sucrose and 5 mM K/Pipes (pH 7) and recentrifuged. The pellet (10-15 mg of protein/g of liver) was resuspended in a small volume of 400 mOSM sucrose and 5 mM K/Pipes (pH 7). "Smooth" and "rough" microsomal fractions were prepared with the Cs' aggregation technique of Dallner (1963). To the 7,800 x g supernatant fraction (see above) was added 1 M CsCl to a final concentration of 15 mM. Eight milliliters of this mixture were layered over 5 ml of 1.3 M sucrose containing 15 mM CsCl. Centrifugation for 135 min at 55,000 rpm (200,000 X gav) in a Beckman type 75 rotor yielded a translucent, reddish pellet ("rough" microsomal fraction) and a well defined band at the gradient interface ("smooth" microsomal fraction). Membranes present at the 0.25 ~/ 1 . 3 M sucrose interface were diluted with 20 volumes of 400 mom sucrose and 5 mM K/Pipes (pH 7). The pellet fraction was resuspended in the same buffer. Both membrane suspensions were recentrifuged and resuspended in a small volume (10-15 mg of protein/ml) of 400 mom sucrose and 5 mM K/Pipes (pH 7). Microsomes were quickly frozen and stored at -65 "C before use. "Intermediate" density rabbit skeletal muscle sarcoplasmic reticulum vesicles used in this study have been characterized previously (Meissner, 1975).
Assays-Protein was estimated by the method of Lowry et al. (1951), using bovine serum albumin as a standard.
Unless otherwise indicated, glucose-6-P and mannose-6-P phosphohydrolase activities were measured at 23 "C in 0.5 ml of a medium containing 50-100 pg of microsomal protein, 50 mM KCI, 2 mM glucose-6-P or mannose-&P, and 5 mM K/Pipes (pH 7.0). Acid-phosphatase activity was measured at 32 "C in 0.5 ml of a medium containing 50-100 pg of microsomal protein, 10 mM 2-glycerophosphate, and 50 mM Na/acetate (pH 5.0). The reactions were started by the addition of the substrate and stopped after 5, 10, and 20 min with 1.0 ml of 1.5 M HC10,. Inorganic phosphate was determined on 1 ml of the proteinfree supernatant using a modification (Rouser and Fleischer, 1967) of the method of Chen and Toribara (1956). Latency of mannose-6-P activity (Arion et al., 1972) was removed by pretreatment of the microsomes (2 mg/ml) at 0 "C with deoxycholate (1.3 mg/ml) in 50 mM KC1 and 5 mM K/Pipes (pH 7). 5'-Nucleotidase activity was measured at 32 "C by the method of Michell and Hawthorne (1965), and succinate-cytochrome c reductase activity was estimated at 32 "C according to Fleischer and Fleischer (1967). All activities were linear with time and protein concentration.
Isotope Flux Measurements-Measurement of radioisotope flux rates involved incubation of liver microsomes with radioactive compounds followed by transfer of the vesicles into a dilution medium. Radioactivity trapped by the vesicles was determined by Millipore filtration. Microsomes were initially transferred to a medium of known composition. Unless otherwise indicated, they were incubated for 4 h at 0 "C in a large volume (0.5-1.0 mg of microsomal protein/ml) of incubation medium (400 mOSM sucrose and 5 mM K/Pipes, pH 7), sedimented by centrifugation for 60 min at 35,000 rpm in a Beckman 42.1 rotor, and resuspended in a small volume (10-15 mg of protein/ m l ) of incubation medium. Radioisotope influx and efflux rates were determined by incubating vesicles from 30 min to 5 h at 23 "C in the presence of radioactive compounds (0.4 mCi/ml of 'H and/or 0.2 mCi/ml of "C). The vesicles were then diluted 500-fold into an unlabeled release medium with rapid mixing. Efflux of the radioactive compounds was terminated by placing aliquots (0.5 ml) on 0.45 pm HAWP Millipore filters followed by rapid rinsing with unlabeled release medium. The time required to execute filtration and rinsing was 15-20 s. The radioactivity retained on the filters was counted in 4.5 ml of a scintillation liquid which completely dissolved the filters. The fluid contained 60 g of naphthalene, 4.2 g of 2,5-diphenyloxazole, 180 mg of 1,4-bis-[2-(5-phenyloxazolyl)]benzene, and 20 ml of water in 900 ml of dioxane. Counting of singly and doubly labeled samples was carried out in a LKB liquid scintillation system, using minivials.
Membrane Potential Measurements-Membrane potentials were generated by gradients of permeant ions between the intravesicular cavity and the medium into which the vesicles were diluted. Membrane potentials (negative inside) were detected by the use of the fluorescent dye diO-Ce(3) (Sims et al., 1974) as previously described (McKinley and Meissner, 1978). Formation of positive membrane potentials and pH gradients did not significantly affect the fluorescence emission of di0-Cs-(3) under the conditions used in the present study (Meissner and Young, 1980). The polarity of membrane potentials is reported according to standard convention, that is, reference (ground) is extravesicular. Fluorescence assays were carried out at 15 "C under stirring in a Farrand model 801 fluorometer. Excitation was at 470 nm and emission was recorded at 495 nm. Slits used resulted in a half-hand width of 2.5 nm. Vesicle concentrations (approximately 25 pg of protein/ml) were used which produced negligible perturbation of the fluorescence emisson during dilution with incubation medium,

RESULTS
Properties of Liver Microsomal Fractions-Electron micrographs of unfractionated rat liver microsomes and the two sucrose gradient fractions revealed closed membranous vesicles with diameters ranging from approximately 0.1-0.2 pm (Fig. 1). Membranes sedimenting through the 1.3 M sucrose layer of the gradient had a rough, granular appearance, suggesting the presence of ribosomes bound to the external, cytoplasmic surface. Membranes collected from the 0.25 M/ 1.3 M sucrose interface of the gradient lacked membranebound ribosomes. The more heterogeneous appearance of the interface fraction reflected the fact that it was likely derived in part from organelles other than smooth endoplasmic reticulum. Enzymatic analysis indicated that unfractionated, rough and smooth microsomes hydrolyzed glucose-6-P, an enzymatic activity typically associated with endoplasmic reticulum of rat liver ( Table I). The fractions contained small quantities of inner mitochondrial membranes, plasma membranes, and lysosomes, as indicated by their low succinatecytochrome c reductase, 5'-nucleotidase, and acid phosphatase activities, respectively.
Zsotope Flux Experiments-The permeability of liver microsomes to ['H]sucrose and ~-['~C]glucose was determined by Millipore filtration as described under "Materials and Methods." The two radioactive compounds (at a concentration of 10 mM) were allowed to move across the microsomal membranes into the vesicles for 1 h at 23 "C. Vesicles were then diluted 500-fold into an unlabeled iso-osmolal dilution medium and were collected on Millipore f i t e r s at time intervals ranging from %-4 min. Radioactivity remaining with the vesicles on the fiters was determined. As shown in Fig. 2, microsomes possessed a permeability barrier for vH]sucrose and D-['4C]glUCOSe. Reasonably straight lines were obtained suggesting that radioisotope efflux may be approximated by fust order kinetics. Comparison of the slopes indicated that the efflux rate of D-glucose (tl,2 = 2-3 min) between 30 s and 4 min was somewhat greater than that of sucrose ( t l r i 2 5 min). A striking difference was that the apparent D-["Cjglucose spaces amounted to only a fraction of the ['H]sucrose spaces. One possible, although not likely, explanation for this difference was that an incubation time of 1 h was sufficient to equilibrate [lH]sucrose but not D-[14CjglUCOSe across the vesicle membranes. We tested for this possibility by incubating microsomes in the presence of [3H]sucrose and ~-[ '~C ] g l u c o s e for times ranging from 30 min to 5 h. Vesicles were then diluted and processed as in Fig. 2. Apparent isotope space for D-glucose was maximal after an incubation period of 30 min (Fig. 3). As expected from the slower efflux rate of sucrose (Fig. 2), a longer time interval (1-2 h) was required to obtain a maximal space for ["Hlsucrose. Extension of the preincubation period to 4 and 5 h resulted in decreased isotope spaces for most but not all preparations. In most cases, data points between 1 and 5 h could be connected on semilogarithmic FIG. 1. Electron micrographs of liver microsomal fractions. Samples were fixed with 2.5% glutaraldehyde and 0.5% tannic acid in 0.3 M sucrose, pH i.2, for 15 min in ice and 45 min at 23 "C and sedimented. Pellets were postfixed with 1% OsO,, embedded, sectioned, and stained as previously described (Malouf and Meissner, 1979). A,B,C, unfractionated, rough, smooth liver microsomal fractions, respectively: X 30,300. plots ( Fig. 3) by two nearly straight and parallel lines suggesting that the decrease in isotope space was of first order and independent of the isotope tested.
A similar difference in sucrose and D-glucose isotope spaces was obtained when 10 mM sucrose or glucose was omitted from the release medium or when incubation and release media were used that contained 400 moSM sucrose or D-glucose (see below). In another control, we found that rabbit skeletal muscle sarcoplasmic reticulum vesicles (Meissner, 1975)  Our explanation for the different isotope spaces of sucrose and D-glucose is that liver microsomes consist of vesicles which differ in their permeability to sucrose and D-glUCOSe. Between 50 and 75% of the vesicles release D-[14C]glUCOSe within 20-30 s, i.e. before the first time point was taken. These vesicles (designated type A) appear, therefore, permeable to glucose, presumably because they contain a mechanism that facilitates the release of D-glucose. The remaining 25-50s of the vesicles (designated type B) had a low D-['4C]glucose permeation rate similar to that of ["H]sucrose and subsequently seem to lack a permeation system for D-glucose.
The apparent isotope spaces of liver microsomes to a variety of additional solutes were determined, as described for ["H]sucrose and D-[14C]glUCOSe in Figs. 2 and 3. The related compounds L-glucose, 2-deoxy-~-ghcose, D-mannOSe, and Dmannitol as well as the nucleoside uridine behaved like Dglucose in that essentially identical efflux rates and apparent isotope spaces were observed. Each of these uncharged solutes had an apparent isotope space of 0.85-1.0 pl/mg of protein which corresponded to 30-40% of the [:'H]sucrose space (Table  11). Similarly low apparent isotope spaces were measured for the univalent cations ""Rb+, "Na', and ["Hlcholine. After 30 s of dilution, the cations left the vesicles with a rate (tIr2 = 1-2 min) about twice as fast as D-glucose (cf. Fig. 2).  chain displayed an isotope space comparable to that of Dglucose. In contrast, isotope spaces of the negatively charged L-glutamate and the positively charged L-lysine were similar to the ["Hlsucrose space. Together with the analogous behavior of the three sugars D-glucose, gluconate-, and glucosa-mine+, these results would seem to suggest that introduction of a positively or negatively charged group into a molecule can retard its movement across type A liver microsomal membranes.
Cl-appeared to pass rapidly the membranes of most vesicles since even after prolonged incubation of the vesicles in the presence of ""Cl-, only small amounts of radioactivity remained with the vesicles 20-30 s after dilution. "'Cl remaining with the vesicles after 30 s of dilution exchanged with external Cl-with a half-time (tllz m 3 min) comparable to that of Dglucose. Apparent ""Cll-impermeable vesicle spaces accounted for about 9, 10, and 4% of the [.'H]sucrose spaces of unfractionated, smooth, and rough microsomes, respectively (Table  I). We interpret these data as indicating that liver microsomes contain a small fraction of vesicles which are quite impermeable to Cl-. The low content of Cl--impermeable vesicles in the rough microsomal fraction suggests that these vesicles were likely derived from organelles other than endoplasmic reticulum.

Osmotic
Swelling and Rupture Experiments-Above we suggested that liver microsomes consist of vesicles, some of which are readily permeable to D-glucose (type A) and some that are not (type B). Both types of vesicles were relatively impermeable to sucrose. One might expect then that dilution of sucrose-filled vesicles into glucose medium will generate an osmotic force in glucose-permeable vesicles due to rapid influx of solute and water. Vesicles will swell and disrupt which in turn would induce rapid release of sucrose thereby lowering internal osmotic pressure of the vesicles. On the other hand, vesicles with a glucose diffusion barrier comparable to that of sucrose should not disrupt and should, therefore, be able to retain their content for longer times. To test this hypothesis, osmotic behavior of vesicles filled with 400 ITIOSM ['HIsucrose or D-['4C]ghCOSe was studied. In control experiments, vesicles were diluted into an unlabeled dilution medium with a com-position identical with the incubation medium and collected a t various time intervals on Millipore fiiters (Fig. 4 ) . ["HI-Sucrose and ~-["C]glucose efflux rates and spaces comparable to those in Fig. 2 were observed. In another control experiment, it was established that both type A and B were osmotically active. Dilution of the vesicles into a medium of low osmolality resulted in release of most, but not all, of the trapped ["HJsucrose and ~-['~C]glucose (Fig. 4). The remaining radioactivity diffused across the vesicle membranes with a rate comparable to that observed under iso-osmolal conditions. Apparently the vesicles reformed a permeability barrier once enough sucrose and glucose was released to lower internal osmotic pressure sufficiently.
A fast and slow efflux component was also distinguished when sucrose-filled vesicles were diluted into iso-osmolal glucose medium (Fig. 4) Table 111). Most of the trapped 'H and 14C radioactivity was also released when vesicles were transferred into a KC1 medium containing the K+-selective ionophore valinomycin. As shown in Fig. 4 for ['HHJsucrose-or D-[14C]glucose-filled vesicles diluted into H20, a rapid initial release phase was followed by normal, slow ['H]sucrose and D-['4C]glUCOSe leakage rates. Since the generation of an osmotic force required the inward movement of both the major cation and anion, it appeared that in the presence of the ionophore, K+ along with C1-and  (not shown). We interpret these results to show that type A vesicles are permeable to D-glucose as well as to K' , Na', Tris', and choline+. Glucose-impermeable (type B) vesicles appeared, on the other hand, to be relatively impermeable to Tris' and choline'. These vesicles thereby avoided massive inflow of Tris+ or choline' along with C1-and water and subsequent rupture. Type B vesicles appeared also to form a permeability barrier to K' and Na+, although at a reduced level in comparison with Tris+ and choline+.
Among the additional compounds listed in Table 111, essentially two groups of solutes could be distinguished. D-mannose, D-mannitOl, uridine, and glycine behaved like D-glucose in that they caused a large initial burst of [:'H]sucrose release. Amounts of D-["c]ghcose remaining with the vesicles were comparable to those found in the D-glucose control medium or were actually greater than in the control medium, as found amounts of ~-['~C]glucose in the presence of the two latter solutes is not clear at present. The second group of components which included K/gluconate (Fig. 4) was intermediately effective in increasing the initial ["H]sucrose leakage rate. The chloride salts of Mg", lysine', and glucosamine+ as well as N-aCetyl-D-glUCOSamine and K/glutamate behaved like K/gluconate in that reduced initial bursts of [:lH]sucrose release were generated on dilution into media containing these compounds (Table 111). Levels of D-['4C]glucose remaining with the vesicles 30 s after dilution were similar to that of the control, Le. when vesicles were diluted into D-glucose medium.
Assuming that influx of salts is largely determined by the slower of the two penetrating ions present in the test medium, the [:3H]sucrose and D-['4C]glUCOSe release experiments suggested significant differences between the permeability barriers formed by type A and B vesicles. Glucose-permeable (type A) vesicles appeared to be readily permeable to a variety of small neutral and charged molecules. Glucose-impermeable (type B) vesicles differed from type A vesicles in that the same solutes, with the exception of K', Na+, and C1-, permeated across their membranes similarly slowly as sucrose. The results of osmotic swelling and rupture experiments were in reasonable agreement with the isotope flux measurements.
Membrane Potential Experiments-Previous isotope and membrane potential measurements have shown that fragmented rabbit skeletal muscle sarcoplasmic reticulum consists of two vesicle populations, one of which contains a K,Na channel and a second one that does not Meissner, 1977, 1978). The ratio of (K',Na')-permeable and -impermeable vesicles was approximately 2:l. Both types of vesicles were relatively impermeable to Ca", Mg", and larger univalent ions such as gluconate-, Pipes-, choline', or Tris'. Fig. 5A illustrates that membrane potential measurements are a useful means to distinguish (K',Na')-permeable and -impermeable vesicles. Sarcoplasmic reticulum vesicles filed with 400 mOSM K/Pipes were diluted into Mg, Ca, Tris, Na, or K/ Pipes media in the presence or absence of the K' ionophore valinomycin. The fluorescent dye 3,3"dipentyl-2,2'-oxacarbocyanine was used to visualize vesicles which could generate and maintain a K'-induced membrane potential (negative inside). Vesicles elicited no change in fluorescence when diluted into K or Na/Pipes medium, indicating that no membrane potential was created (McKinley and Meissner, 1978). However, when vesicles were diluted into Mg, Ca, or Tris/ Pipes medium in the absence of valinomycin, fluorescence signals of intermediate size were detected. Under these conditions, only the (K',Na')-permeable vesicle fraction formed a negative membrane potential. The absence of a potential in (K',Na')-impermeable vesicles was due to the similarly low permeability of K', Mg", Ca", and Tris'. Addition of valinomycin to the Mg", Ca", or Tris' medium made (K',Na')impermeable vesicles permeable to K' and resulted in a maximal dye signal since it enabled formation of a membrane potential in the entire vesicle population, i.e. (K+,Na')-permeable and -impermeable vesicles. On the other hand, a negative potential was exclusively formed in (K',Na')-impermeable vesicles by transferring the entire sarcoplasmic reticulum vesicle population from K' to Na' medium containing valinomycin. Under these conditions, no membrane potential was formed in (K+,Na')-permeable vesicles. Because of the presence of the K,Na channel, these vesicles rapidly exchanged all of their K' for Na' within 1-2 s, the experimental limit of detection. Within 1 h, fluorescence signals returned to their original values. This gradual reduction in membrane potential was likely due to slow inward movement of the extravesicular cation (Mg", Ca2', Tris', or Na') and the eventual dissipation of the K' gradients.
Liver microsomes showed a similar, although not identical, behavior when transferred from K/Pipes to Mg or Ca/Pipes medium in the presence or absence of valinomycin (Fig. 5B). Pipes-, a relatively slowly penetrating anion (cfi Table 111) was used to minimize osmotic effects. As in sarcoplasmic reticulum, dye responses of different magnitude were observed in the presence and absence of valinomycin, suggesting the presence of two liver microsomal vesicle populations which differed in their K' permeability. Decrease in fluorescence in the absence of valinomycin pointed to the presence of vesicles (type A) which were more permeable to K' than to Mg'+ or Ca" and which, therefore, formed a negative membrane potential. Increase in fluorescence signal in the presence of valinomycin suggested the presence of a second population of vesicles (type B) which had in the absence of the K ionophore a K' permeability similar to that of Mg'+ or Ca". The signal seen in the presence of valinomycin represented the entire vesicle population, i.e. type A and type B microsomes. In support of our interpretation was that the ratio of fluorescence decreases and apparent isotope spaces assigned to type A and type B vesicles were about the same. Liver microsomes also generated a fluorescent signal of intermediate size when transferred from Na or Tris to Mg or Ca/Pipes medium (not shown). Thus, in agreement with the osmotic rupture experiments (cf . Table III), fluorescent measurements indicated that a substantial portion of the microsomes was more permeable to K' , Na+, and Tris+ than to Mg" or Ca". Liver microsomes and sarcoplasmic reticulum vesicles differed in their K' permeability in that the initial fluorescence B LM In K Pipes r w 5 min were diluted 60-fold into an iso-osmolal Pipes medium containing the indicated cation, 1.5 FM diO-C~(3), and no or 0.5 PM valinomycin (Val). The K' of the vesicle medium served to establish an initial 60fold K' gradient throughout the experiments. changes in Mg or Ca/Pipes medium were appreciably slower for liver microsomes (20-30 s) than for sarcoplasmic reticulum vesicles (-2 s). Despite their preferential K' permeability, a majority of liver microsomes appeared, therefore, to lack an ion-conducting structure for K' such as the K,Na channel which renders two-thirds of sarcoplasmic reticulum vesicles highly permeable to K' .
Absence of a "channel" rendering liver microsomal vesicles highly permeable to K' and Na' was supported by the following two observations. First, transfer of microsomes from K' to Tris' medium did not elicit an appreciable dye response, suggesting that most of the microsomes did not form a potential. Second, liver microsomes rendered permeable to K+ by valinomycin elicited a similar dye signal when transferred to Na' or Tris' medium. The rapid collapse of membrane potentials in Na' or Tris' medium as well as the reduced magnitude of dye signals in Na' or Tris+ medium, as compared to those in Mg2+ or Ca" medium, were in accord with the differential permeability of liver microsomes to these cations (cb Table  111).
In summary, membrane potential experiments support our contention that liver microsomes are more permeable to K' , Na+, and Tris' than to Mg2+ or Ca2'. However, the majority of liver microsomes appeared to lack a highly conducting K' and Na' structure, such as the K,Na channel of sarcoplasmic reticulum.
Permeability of Liver Microsomes to Glucose-6-P-In osmotic swelling and rupture experiments, liver microsomes were routinely kept for 90 min at 23 "C in order to allow ["Hlsucrose and D-['4C]glucose to diffuse across vesicle membranes. Table IV shows that latency of mannose-6-P phosphohydrolase activity (Arion et al., 1972) was maintained, suggesting that liver microsomes retained a permeability barrier to mannose-6-P. In addition, nearly identical glucose-6-P phosphohydrolase activities by intact or disrupted microsomes before and after incubation indicated that glucose-6-P

Kinetic parameters and effects of reagents on glucose-6-P and mannose-6-P phosphohydrolase activities of liver microsomes
Experiment 1: unfractionated liver microsomes (10 mg of protein/ ml) present in 5 mM K/Pipes buffer (pH 7) containing 400 moSM sucrose or glucose were pretreated by incubation (i) for 90 min at 23 "C, or (ii) with 1 mM DIDS for 5 min at 30 "C. Treated samples were then either directly diluted into the standard glucose-6-P phosphohydrolase assay medium (cf. "Materials and Methods") or were first disrupted before the assay by transferring them into standard glucose-6-P phosphohydrolase assay medium containing microsomes at a protein concentration of 2 mg/ml and 0.13% deoxycholate (DOC). Data are the average of four determinations. S.E. = &IO%. Experiment 2: kinetic parameters of glucose-6-P and mannose-6-P phosphohydrolase activities were determined from Lineweaver-Burk plots. Dependence of the enzymatic activities of intact and deoxycholatetreated (see above) liver microsomes on their respective substrate concentrations (0.5-10 mM) was determined at 23 "C in a 5 mM K/ Pipes buffer (pH 7) containing 400 m o m KCI. transport and phosphohydrolase activities were not appreciably impaired by preincubation for 90 min a t 23 "C. Glucose-6-P phosphohydrolase activity by intact microsomes was inhibited by the anion transport inhibitor DIDS, as previously reported (Zoccoli and Karnovsky, 1980) (Table IV). Disruption of the permeability barrier by treating membranes with deoxycholate reversed inhibition of glucose-6-P phosphohydrolase activity. These results were in accord with previous suggestions that blockage of glucose-6-P permeation will limit glucose-6-P access to its hydrolytic site on the cisternal membrane surface (Arion et al., 1972;Gold and Widnell, 1976;Ballas and Arion, 1977;Zoccoli and Karnovsky, 1980). Kinetic parameters of glucose-6-P phosphohydrolase in intact and deoxycholate treated microsomes were determined under conditions similar to those used in isotope flux experiments using varying concentrations of glucosn-6-i-(0.5-10 mM). The double reciprocal Lineweaver-Burk plots 01 the data yielded two straight lines (not shown). From the intercepts with the abscissa and ordinate, apparent K,,, values of 6.1 and 2.9 mM and V,,, values of 0.10 and 0.15 ymollmg of protein/min were obtained for intact and detergent-disrupted microsomes, respectively. Corresponding K,,, and v,,, values for mannose-6-P hydrolase activity by deoxycholate treated microsomes were 6.1 mM and 0.15 pmol/mg of protein/min, respectively (Table  IV). The increase in V,,,, from 0.10-0.15 mol/mg of protein/ min raised the possibility that one-third of the intact microsomes were not capable of glucose-6-P hydrolysis because of their lack of an efficient glucose-6-P permeation system. An alternative explanation would be that a change in membrane structure during detergent treatment resulted in an increased turnover rate of glucose-6-P hydrolysis.
A direct test for the presence of a specific glucose-6-P permease involved dilution of microsomes filed with 390 mOSM ["H]sucrose and 10 mOSM D-[' 4C]glUCOSe or fiiled with 400 mOSM D-['4C]ghCoSe into media containing valinomycin and the K salts of glucose-6-P, mannose-6-P, glucose-1-P, or uridine-5'-monophosphate. Transfer of the vesicles to glucose-6-P medium increased ['H]sucrose leakage within the first 3 0 s, resulting in a reduction of the [:'H]sucrose space by 20-25": (Fig. 4, Table V). [:'H]Sucrose space was similarly decreased when vesicles were diluted into media containing the impermeant anion mannose-6-P (Ballas and Arion, 1977), glucose-1-P, or uridine monophosphate (Table V). Moreover, nearly complete blockage of glucose-6-P permeation by DIDS (cf Table IV) did not significantly reduce [:'H]sucrose leakage on transfer of vesicles into glucose-6-P or glucose-1-P media. In another control, a similar reduction in [:1H]sucrose space was observed within the fist 30 s when sarcoplasmic reticulum vesicles were diluted from ["H]sucrose medium to K/glucose-6-P medium containing valinomycin (not shown). The K salts of the above monophosphates were also ineffective in increasing the initial leakage rates of D-['4C]glUCOSe from liver microsomes. Together, the above results suggest that liver microsomal membranes are appreciably less permeable to glucose-6-P than to D-ghC0Se.
Transfer of liver microsomes from sucrose medium to media containing valinomycin and the K salts of phosphate, sulfate, or oxalate reduced the [3H]sucrose space by about one-half within 30 s (Table V). In type A vesicles, phosphate, sulfate, and oxalate anions appeared, therefore, to pass the membrane with a rate intermediate between those of sucrose and Dglucose. Initial leakage rates of D-['4C]gl~~ose were only slightly increased, suggesting that the divalent anions crossed type B vesicle membranes with a comparatively slow rate.
A significant reduction in the initial burst of [:'H]sucrose and an increase in the apparent D-['4C]ghcose vesicle space were seen on transfer of DIDS-treated microsomes into D-

Apparent [3H]sucrose and D-['4C]glUCOSe spaces of untreated and DIDS-treated liver microsomes diluted into media containing the K salts of glucose-6-P and other anions
Rough liver microsomes were incubated in 5 mM K/Pipes buffer (pH 7) containing 390 m o m [JH]sucrose and 10 m o m D-['4C]glucose for 90 min at 23 "C followed by incubation for 5 min at 30 "C in the presence or absence of 1 mM DIDS. The microsomes were then diluted 500-fold at 23 "C into 5 mM K/Pipes buffer (pH 7) containing the indicated solute and no or 0.5 PM valinomycin (Val). Isotope spaces obtained 30 s after dilution of microsomes into the test media are given. Data are the average of three or more determinations. S.E. = &20% or less.

Composition of dilution medium
Sucrose o-Glucose o-Mannitol KC1 (+Val) Choline/Cl Glucose-6-P (K + Val) Mannose-6-P (K + Val) Glucose-1-P (K + Val) Uridine-6-P (K + Val) The permeability of rat liver microsomes has been previously investigated. Nilsson et al. (1973) found that uncharged solutes with a molecular weight of up to 600 including Dglucose and sucrose equilibrated across the microsomal membranes within 2 h, the time used to pellet microsomes by centrifugation. By contrast, microsomes appeared to be impermeable to charged molecules with a molecular weight as low as 90. Ballas and Arion (1977) used the technique of centrifugal transfer through a layer of silicone oil to separate 30 r -< : 10 -0 3 -["H]sucrose leakage rate was also reduced when DlDS-treated microsomes were diluted into media containing D-mannitol, N-acetyl-D-glucosamine, or valinomycin plus the potassium salts of glutamate or phosphate (Table V). By contrast, no or only relatively small changes in ['H]sucrose and ~-['~C]glucose vesicle spaces were noted in media containing lysine/ HC1, choline/Cl, or KC1 plus valinomycin. The observed increase in sucrose and D-glucose isotope spaces raised the possibility that treatment with DIDS significantly lowered the permeability of liver microsomes to certain uncharged and negatively charged solutes such as D-glucose and gluconate-. glucose or K/gluconate medium (Fig. 6, Table V). A small, but consistent, reduction in the permeability of the microsomes t o sucrose was also observed on treatment with 1 mM DIDS. Use of 5 mM DIDS resulted in an appreciable loss of [''H]sucrose and D-['*c]glUCOSe spaces, suggesting a general breakdown of the permeability barrier (not shown). The initial = 4000). Mannose-6-P was also found to penetrate the microsomal membranes, although at a slower rate than glucose-6-P. Our studies indicated that the majority of microsomes were appreciably less permeable to glucose-6-P and mannose-6-P than to D-glucose or D-mannose. Evidence for low permeability of intact microsomes to larger anions in general was previously provided by the observation that EDTA solubilized only a small portion of lead phosphate precipitates formed within the microsomal vesicle space by incubation in the presence of glucose-6-P and Pb2+ (Gold and Widnell, 1976). The previous studies did not describe the presence of two types of vesicles which differ in their permeability to small solutes.
The most economical explanation for the permeability of type A vesicles to various solutes would be that these vesicles contain a single permease or "channel" which selects against solutes on the basis of their size and charge. The three uncharged solutes D-glucose, N-acetyl-D-glucosamine, and sucrose passed with a decreasing permeation rate across the membranes of type A microsomes. Taking into consideration the van der Waal's dimensions of these 3 molecules, solute selectivity of liver microsomes can be explained in geometrical terms by assuming that type A vesicles contain a channel with a diameter of approximately 7-8 A. Accordingly, solutes with a minimal cross-section of less than 7-8 8, such as D-glucose would be able to pass through the pore, while solutes with a cross-section exceeding 7-8 A such as sucrose would be retarded. Challenging this hypothesis was our observation that the permeability of type A membranes to both glucosamine' and gluconate+ was lower than that to D-glucose despite the fact that both molecules are small enough to fit through a pore with a diameter of 7-8 A. It seems, therefore, that, in addition to size, overall charge is important in determining the permeation rate of a molecule through the channel. Selectivity against charged molecules seemed to become more pronounced with a n increase in charge density, as evidenced by the limited permeability of type A microsomes to the relatively small divalent ions Mg'+ , sulfate", or oxalate2-. In support of a single channel structure was that DIDS treatment decreased permeability of type A microsomes to several solutes including D-glucose and gluconate-. At present, it cannot be ruled out, however, that liver microsomes contain more than one permeation system. In fact, DIDS treatment did not significantly reduce membrane permeability to positively charged molecules such as choline+ or lysine'. Enzyme competition (BaLlas and Arion, 1977) and DIDS labeling (Zoccoli and Karnovsky, 1980) studies have raised the possibility that distinct permeation systems exist for glucose-6-P, Pi, and P,P,. Another argument against a one-channel model with a wide substrate specificity would be that liver microsomes are able to distinguish between the two closely related compounds glucose-6-P and mannose-6-P (Ballas and Arion, 1977; Table   IV). In part, we were not able to distinguish between a multiand one-channel permeation system because our experimental approaches were limited. For example, we could not determine the exact permeation ratio of two fast or slow moving solutes such as D-and L-glucose or glucose-6-P and mannose-6-P. In agreement with previous reports (Ballas and Arion, 1977) enzymatic analysis indicated a faster permeation rate for glucose-6-P than mannose-6-P. The difference in the rates was seen because, in contrast to osmotic rupture experiments, enzyme assays could be carried out at a substrate concentration below the K , of the glucose-6-P permease thereby minimizing nonmediated permeation. On the other hand, we could clearly distinguish between readily and slowly permeating solutes such as D-glucose and sucrose. Estimating liver microsomes to have on an average a diameter of 0.15 pm (Fig. 1) and assuming sucrose permeation to be a fist order process with a half-time of 5 min (Fig. 2), the sucrose permeability coefficient is calculated to be on the order of lo-* cm/s. For comparison, from enzymatic data of Table IV, a glucose-6-P permeability coefficient of about lo-" cm/s is calculated, assuming a substrate concentration of 0.16 M and glucose-6-P permeation to be the rate-limiting step. Minimal ["Hlsucrose leakage from sucrose-fied vesicles on transfer into glucose-6-P medium also suggested a similarly low permeation rate for glucose-6-P and sucrose. By contrast, complete release of D-['4C]glucose from type A vesicles within 20-30 s suggested that these vesicles were at least 20 times more permeable to glucose than to glucose-6-P or sucrose. Isotope flux, osmotic rupture, and membrane potential measurements all indicated that solute permeation was likely free rather than coupled to another solute.
Glucose-permeable, type A microsomes are not considered to be simply "leaky" vesicles since both types of vesicles formed a permeability barrier to sucrose, Mg", and Pipes-.
The presence of type A vesicles in rough and smooth liver microsomal fractions further suggests that permeases mediating the movement of various solutes are distributed over the entire reticulum structure. The existence of vesicles differing in their permeability properties in isolated membrane fractions may be explained by assuming that the in vivo reticulum structure contains a limited number of permeable structures. Fragmentation of the reticulum during homogenization will, therefore, yield some vesicles which contain a channel and others that do not. In this regard, it is of interest that (K',Na+)-permeable and -impermeable vesicles with a ratio of about 2:l have been isolated from skeletal muscle sarcoplasmic reticulum. Sonication and reconstitution experiments have led us to suggest that formation of (K+,Na')-impermeable sarcoplasmic reticulum vesicles is due to a limited number of randomly distributed K,Na channels, approximately 50/ pm2 (McKinley and Meissner, 1978;Young et al., 1981).
Despite the presence of (K+,Na')-permeable and -impermeable vesicles in both the rat liver and rabbit skeletal muscle microsomal fractions, several important differences were noted to exist between the cation permeability of the two membranes. The majority of sarcoplasmic reticulum vesicles appeared to be relatively impermeable to D-glucose as indicated by the similar D-ghCOSe and sucrose isotope spaces and efflux rates. This would suggest that skeletal muscle sarcoplasmic reticulum either lacks or contains only a small number of D-glucose-permeable structures. By contrast, rat liver endoplasmic reticulum seems to lack a substantial number of structures which would render tbis membrane highly permeable to K' and Na+, as is the case with sarcoplasmic reticulum. Another significant difference between these two membranes is that sarcoplasmic reticulum membranes contain a H'permeable pathway, whereas liver microsomes appear to lack a pathway that renders them highly permeable to H' (Meissner and Young, 1980). Different permeability behavior of liver and skeletal muscle reticulum membranes is in accord with their different physiological function. Sarcoplasmic reticulum controls skeletal muscle contraction and relaxation by rapidly releasing and sequestering Ca'+. Appreciable Ca'+ fluxes are possible because they can be compensated in charge by the counter movement of H', K+, and Na' via the H' and K,Na channels of sarcoplasmic reticulum (Meissner, 1981). Because of the absence of highly active Ca'+-releasing or -sequestering transport systems in liver endoplasmic reticulum, specific high capacity permeation systems for H', K', and Na' appear not to be required for this membrane. A major function of endoplasmic reticulum of liver is the biosynthesis of lipids and proteins. An NH2-terminal extension of about 20 relatively hydrophobic amino acids is cleaved from the growing polypeptide chain after transfer across the reticulum membrane. While the fate of this "leader" sequence is not known, its degradation inside the endoplasmic reticulum cisternae is conceivable. Similarly, glucose residues are released during oligosaccharide processing inside the reticulum cisternae. Glucose-6-P hydrolysis by endoplasmic reticulum requires that glucose-6-P, P,, and glucose cross the reticulum membrane.
One possible function of the channel(s) present in type A vesicles is to facilitate these processes by mediating the rapid inward and outward movement of substrates and reactants. Whether the D-glucose-permeable channel has any direct role in directing the growing polypeptide chain or oligosaccharide precursors across the reticulum membrane remains to be established.