THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1989 by The American Society for Biochemistry and Molecular Biology, Inc

. 264, NO. 3, Issue of ‘January 25, pp. 1735-1741, 1989 Printed in U.S.A.

Epidermal Growth Factor Receptor of the Intestinal Enterocyte LOCALIZATION TO LATEROBASAL BUT NOT BRUSH BORDER MEMBRANE*

(Received for publication, July 15, 1988)

Lawrence A. Scheving, Robert A. ShiurbaS, Toan D. NguyenQ, and Gary M. Gray From the Departments of Medicine (Gastroenterology Division) and $Pathology and the Stanford Digestive Disease Center, Stanford University School of Medicine, Stanford, California 94305

Interaction of epidermal growth factor (EGF) with its specific receptor (EGFR) was explored in the intact rat small intestine and in highly purified isolated en- terocyte membrane preparations. Despite the fact that the EGF ligand is known to be present at physiological concentrations within the intestinal cavity, no signifi- cant binding of the ligand to the brush border surface was observed. Instead, binding of EGF to the EGFR was confined to other membrane populations, and cor- relation of ligand interaction with the laterobasal membranes (LBM) was nearly perfect ( p < 0.001) across a special equilibrium gradient enriched in brush border and LBM but devoid of intracellular mem- branes. Specific binding to another minor population of intracellular membranes that migrated to a position less dense than typical endoplasmic reticulum-Golgi vesicles on equilibrium gradients was also observed. Immunocytochemical exposure of intestine to EGFR antibody confirmed the localization of the EGFR to LBM and intracellular membranes. As estimated from the intensity of the staining, there may be immunolog- ically active but nonbinding receptor species in the intracellular membrane compartment. Thus, despite the secretion of EGF into the intestinal lumen, the growth and maturational effects of EGF probably re- sult from a specific interaction between EGF and EGFR solely at the laterobasal surface of the entero- cyte. The functional role of the intracellular membrane species of EGFR, which remains to be established, may involve a source of inactive receptor that can be rap- idly recruited and transferred to the LBM surface un- der changing environmental conditions.

Epidermal growth factor (EGF),’ a 6-kDa peptide, binds to a single-chain transmembrane glycoprotein receptor and in- duces growth and differentiation of epithelial and mesenchy- mal cells. Named originally for its stimulatory effects on proliferation and keratinization of epidermis (Carpenter and Cohen, 1979), EGF has since been shown to have actions on many other cell types (Scheving et al., 1980; Yeh et al., 1981),

* This work was supported by National Institutes of Health Na- tional Research Service Award Fellowship DK07904 (to L. A. S.) and by National Institutes of Health Grants DK11270 and DK15802 (to G. M. G.). 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.

NC 27705. J Present address: Dept. of Medicine, Duke University, Durham,

The abbreviations used are: EGF, epidermal growth factor; EGFR(s), epidermal growth factor receptor(s); LBM, laterobasal membranes; BB, brush border; Hepes, 4-(2-hydroxyethyl)-l-pipera- zineethanesulfonic acid; ER, endoplasmic reticulum.

yet its biological role in the small intestine has only recently been the subject of active investigation (Chabot et al., 1983; Dembinski et al., 1982; Scheving et al., 1980).

Although specific cell surface EGF receptors have been identified by radioligand binding studies on isolated cells from intestinal crypts and villi (Forgue-Lafitte et al., 1980; Gallo- Payet and Hugon, 1985), the subcellular location of intestinal EGFR in situ is uncertain. Enterocytes of the small intestine are morphologically and functionally polarized columnar cells possessing distinct laterobasal (LBM) and brush border (BB) surface membranes. Each surface membrane has access to EGF, the LBM surface being continuously exposed to EGF circulating in the blood (Carpenter and Cohen, 1979) and the BB being directly in contact with EGF secreted by Brunner’s duodenal glands, crypt Paneth cells, and goblet cells (Heitz et al., 1978; Poulsen et al., 1986). Because isolated villus cells bind EGF, it seemed likely that they would preferentially display EGF receptors on their brush border surface. Indeed, receptors for EGF were recently found in partially purified BB, LBM, and microsomal intestinal membrane preparations (Gallo-Payet and Hugon, 1985). However, the effects of in- traintestinally infused exogenous EGF on intestinal growth and function are controversial (Dembinski et al., 1982; Fitzpatrick et al., 1987; Goodlad et al., 1987; Skov et al., 1984; Ulshen et al., 1986).

By use of biochemical and immunocytochemical ap- proaches, we have localized the EGF receptor in the rat small intestine. Although the EGF receptor was present on villus cells, no significant binding of EGF to purified BB membranes of the polarized enterocyte could be detected. Specific receptor interaction with radiolabeled EGF was confined to the LBM and to a special population of intracellular membranes.

EXPERIMENTAL PROCEDURES

Enzyme and Protein Assays-Enzymes established to be highly enriched in particular enterocyte membrane fractions were used as markers for these fractions. Mannosidase I1 (for Golgi), potassium- stimulatedphosphatase (for LBM), and sucrase (for BB) were assayed as previously described (Ahnen et al., 1982; Cezard et al., 1979; Nguyen et al., 1987; Tulsiani et al., 1982).

Membrane Fractionation-Male Sprague-Dawley rats (250-300 g) maintained on Purina Rat Chow and exposed to alternate 12-h periods of a light and dark environment were provided with food and water ad libitum. Three h before the onset of the dark period, prior to the usual circadian feeding phase, animals were anesthetized with either diethyl ether or CO, for 2 min and then killed by decapitation.

of T - E G F (150-170 Ci/mmol, Du Pont-New England Nuclear) alone In some experiments, rats were injected intraperitoneally with 2 pCi

or in combination with a 1000 M excess of unlabeled EGF (Bethesda Research Laboratories) at 2 or 10 min before killing. Segments of jejunum and proximal ileum (about 20 cm in length) were quickly excised, rinsed once with cold 1.4% NaCl, and placed on ice. All subsequent steps in membrane purification took place at 4 ‘C. Intes- tinal segments were rinsed via a Pasteur pipette with 1.4% NaCl.

Segments of intestine were everted over a 15-cm blunt-ended glass

1735

1736 Location of Intestinal Epidermal Growth Factor Receptor

rod, and the enterocytes were either scraped free from the muscularis externa using a glass slide or dispersed and harvested as described by Nguyen et al. (1987). In initial experiments, the mucosa from 60-cm jejunal segments was scraped directly into a Waring Blender contain- ing 80 ml of binding buffer (20 mM HEPES, pH 7.4) and homogenized for 45 s at the highest setting. Membrane vesicles could thus be prepared within 5 min after killing in the absence of EDTA. Subse- quently, the vesicles were centrifuged at 60,000 X g for 20 min to remove cytosol. Phase-contrast microscopic examination revealed the absence of intact cells and the presence of membrane fragments, including some intact nuclei and brush borders. The pellet from the initial centrifugation was resuspended in 6 ml of binding buffer and divided into two equal 3-ml aliquots for incubation with 5 X lo-" M "'I-EGF alone or with a 1000 M excess of unlabeled EGF (for nonspecific binding). The membranes were then layered directly onto a 25-60% sorbitol gradient containing binding buffer and centrifuged in an SW 27 rotor at 85,000 X g for 16 h to equilibrium. Fractions (2 ml each) were collected with a Buchler Auto Densi Flow I1 C instru- ment and the radioactivity determined. Assays for enzymes known to be enriched in various membrane fractions were performed on all fractions (cf. enzyme assays above).

In other experiments, comprehensive separation of enterocyte membrane populations was carried out as previously detailed in this laboratory by Ahnen et al. (1982) and Nguyen et al. (1987). Apprecia- ble enrichment of BB (15-fold), LBM (15-fold), and of a combined endoplasmic reticulum (ER)-Golgi (10-fold) was achieved. Highly purified LBM, free of contamination with ER and Golgi membranes, was also prepared (Nguyen et al., 1987). These purified membranes were examined for their capacity to interact specifically with EGF.

EGF Binding Studies-For binding studies, up to 200 pg of mem- brane protein from the various subcellular fractions were incubated with "'I-EGF (final concentration 5.67 X lo-" M) in 0.5 ml of 20 mM Hepes buffer, pH 7.5, containing 0.1% bovine serum albumin and 50 mM MgClz in a 12 X 75-mm polypropylene tube for 1 h at 22 'C. This buffer had been centrifuged at 38,000 X g for 20 min and prefiltered through a Nalgene 0.20-pm membrane prior to use. After the incu- bation, the bound EGF was separated from the unbound by centri- fugation at 38,000 X g for 15 min. Binding of EGF to the tube was less than 1% of the initial amount added. Nonspecific binding of lz5I- EGF, measured in the presence of 5 X lo-' M unlabeled EGF, was 1% of the total binding. Specific binding was calculated by subtracting nonspecific binding from total binding. Kinetics of the EGF binding to membranes was derived by plotting the data by the method of Scatchard (1949).

Immunocytochemistry-Segments of duodenum, jejunum, and ileum were rapidly removed and freeze-substituted as described (Alqu- ist, 1972; Michael et al., 1984). Briefly, tissues were quick-frozen in chlorodifluoromethane (Freon 22) at -150 "C, transferred to a solu- tion of ch1oroform:methanol:acetone (1:2:1) at -80 "C, and main- tained in this solution for 2 weeks. After warming to 4 "C, the freeze- substituted tissues were washed in cold chloroform before infiltration with paraffin and embedding into blocks. The tissue was then cut into 5-pm sections and mounted onto glass slides for immunoperox- idase (Sternberger, 1979) or immunogold staining.

Paraffin-embedded tissue sections were processed for immuno- staining for the EGF receptor by modification of the immunogold- silver staining method of Holgate et al. (1983) and Danscher and Norgaard (1983). Sections were de-waxed in xylene, rehydrated in graded ethanol series (100->70%), and soaked in Tris-buffered saline (0.05 M Tris, 0.15 M NaCI, pH 7.6). Lugol's iodine solution (0.05% (w/v) iodine crystals in absolute ethanol) was applied for 5 min, and removed by thorough rinsing in deionized water, followed by treat- ment in 2.5% (w/v) sodium thiosulfate for 3 min. After a brief rinse in water, slides were pretreated for 60 min in 0.05 M Tris, 0.1 M glycine, 0.1 M NH4CI, pH 7.6. Five antibodies were then used. The primary antibody, mouse monoclonal 29.1.1 IgG (from ICN ImmunoBiologicals), was diluted 1:5000 in Tris-buffered saline and applied for 18 h at room temperature. This well characterized mono- clonal, originally raised against the human EGF receptor in A431 epidermoid carcinoma cells, binds with high affinity to an extracel- lular portion of the receptor distinct from the EGF binding domain on both human and rodent cells. A well characterized rabbit poly- clonal antibody against mouse epidermal growth factor receptor (gen- erously provided by Eileen D. Adamson, Cancer Research Center, La Jolla Cancer Research Foundation) was applied to different slides at the same dilution for the same length of time. This antibody has comparable affinity for both rat and mouse EGF receptor (Weller et al., 1987). Parallel experiments were conducted using a rabbit poly-

clonal monospecific antiserum to brush border rat sucrase and amino-oligopeptidase (both previously characterized in our labora- tory) (Ahnen et al., 1982; Cezard et al., 1979) and to submandibular gland mouse epidermal growth factor (from ICN ImmunoBiologicals).

Sections were washed for 30 min in Tris-buffered saline (50 mM Tris, 150 mM NaC1, pH 7.6) to remove excess antibodies. Colloidal gold (5-nm particles, Janssen Pharmaceutica, Piscataway, NJ) ad- sorbed to rabbit anti-mouse IgG (for EGF receptor monoclonal anti- body) or goat anti-rabbit IgG (for EGF receptor, sucrase, and EGF polyclonal antibody) was applied overnight at high dilution. Anti- body-conjugated gold particles were diluted in Tris-buffered saline containing0.5% (v/v) cold-water fish skin gelatin to an optical density of 0.005 at 520 nm (the wavelength of maximum light scattering for 5-nm colloidal gold) and applied to tissue sections for 24 h.

After preincubation for 30 min in Tris-buffered saline followed by 30 min in deionized water, the antibody-gold complexes specifically bound to the tissue sections were visualized by gold-catalyzed chem- ical reduction of silver ions to black metallic silver grains. In a photographic darkroom under safelight, tissue sections were im- mersed for 10 min in a silver lactate/hydroquinone physical developer solution containing 20% (w/v) purified gum arabic (Spectrum Chem- ical Mfg. Corp., Gardena, CA) in citrate buffer, pH 3.5. Silver- enhanced tissue sections were washed with six changes of deionized water before exposure to daylight, counterstained with hematoxylin, dehydrated in ethanol, and mounted with microscopic coverslips. A positive immunostaining reaction was observed by light microscopic visualization of intracellular discrete black metallic silver grains, each of which contains a single catalytic gold particle a t its core.

Negative controls included similarly processed control slides in which primary antibodies were either omitted or replaced with mye- loma IgG1, nonimmune mouse ascites, or nonimmune rabbit serum. Positive controls included staining of similarly treated sections of liver, which are estimated to have 20-30 times as many EGF receptors as compared to intestine. Prior to use, specificity for each antibody was confirmed by sodium dodecyl sulfate-polyacrylamide gel electro- phoresis gel-immunoblot analysis (Weller et al., 1987) of tissue ex- tracts rich in the particular antigen of interest (A431 epidermoid carcinoma cell extract for monoclonal and polyclonal antibodies against the EGF receptor, enterocyte brush border extract for sucrase and amino-oligopeptidase, and rat submandibular gland extract for EGF) .

RESULTS

EGF Binding to Everted Gut Sacs and Dispersed Entero- cytes-Previous experiments with crude membrane prepara- tions suggested the presence of specific EGF binding on both BB and LBM membranes (Chabot et al., 1983; Thompson, 1988). We compared the binding characteristics at the apical and laterobasal surfaces in intact enterocytes by the use of everted intestinal segments whose BB surface is exposed to ligands in the surrounding buffer. The intestine (12-cm length) was everted over a glass rod, tied at both ends, and exposed to a lo-" M lz5I-EGF at 4 "C for 45-90 min. There was considerable variation in results due to relatively high (-50%) but variable (&25%) nonspecific binding. Similar low specific binding (1-3%) of EGF to everted sacs was noted in 16 adjacent intestinal segments from duodenum to jejunum. When mucosal scrapings from the gut sac experiments were rapidly homogenized (80 ml of 20 mM Hepes (pH 7.4) in a Waring Blendor a t maximal speed for 40 s), centrifuged to remove cytosol, and the membranes fractionated by sorbitol equilibrium centrifugation, no significant "'I-EGF binding activity was associated with BB membranes. Instead, the ligand was associated with LBM (data not shown). Thus, the apparent EGF binding to the everted sac probably represented association of the ligand with laterobasal membranes through gaps between adjacent enterocytes in the gut sac.

To examine the capacity of the LBM surface to bind EGF, we used two approaches to disrupt tight junctions between enterocytes, thereby exposing the LBM to the T - E G F probe. When everted segments were exposed to 37 "C for 15 min after a 90-min preincubation at 4 "C or the cells were disso-

Location of Intestinal Epidermal Growth Factor Receptor 1737

ciated by buffered EDTA (see under “Experimental Proce- dures”), the specific lZ5I binding increased to -15%. These results suggested that the vast majority of active EGF cell surface receptors resides on the LBM of the enterocyte and prompted us to define more precisely the EGFR localization with highly purified membrane preparations.

Properties of EGF Binding-Preliminary experiments with lZ5I-EGF and purified intestinal LBM revealed that maximal specific binding occurred at 22 “C after 50 min of incubation and remained stable for at least 90 min (data not shown), and binding experiments were carried out at 22 “C for 60 min in the subsequent studies detailed below. Specific binding of EGF was a linear function of the amount of membrane protein added from 30 to 300 kg/ml of assay, and nonspecific binding was less than 1% of the total binding. Specific binding was consistently 2-3 times higher when conducted with mem- branes in 20 mM Hepes buffer, pH 7.5, compared with either a 200 mM sodium-potassium buffer or Ringer’s solution at the same pH. The addition of 5 mM MgClz or CaC12 enhanced EGF binding in membranes prepared with EDTA buffers and therefore was included in binding studies with membranes that had been exposed to the chelator. The K d was determined by the method of Scatchard to be of high affinity, -10”’ M lZ5I-EGF (data not shown).

Localization of the EGF Receptor in Rapidly Prepared En- terocyte Membrane Fractions-Because of the possibility that EGF receptors might migrate in the plane of the surface membrane between the BB and LBM during incubation of everted sacs or dispersed enterocytes, we carried out rapid mucosal scraping (60 cm of intestine over 3 min) at 4 “C, cell lysis in 20 mM Hepes in a Waring Blender at maximal setting for 45 s, and centrifugation (50,000 x g, 30 min) followed by resuspension of the crude membranes in binding buffer with lo-“ M lZ51-EGF alone or with M unlabeled EGF to allow estimation of nonspecific binding. The membrane populations were then separated by equilibrium centrifugation in a 25- 60% sorbitol gradient (see under “Experimental Procedures”) and the collected gradient fractions analyzed for radioactivity and for the appropriate membrane marker enzymes (see under “Experimental Procedures”).

As shown in Fig. 1, the specific binding of EGF was maximal in the fractions enriched in ER-Golgi and LBM, but there was also a broad distribution extending into the lower density region (fractions 5-10, Fig. 1). Notably, no peak was seen in the region of the brush border membranes. Also, when lZ5I- EGF was bound to dispersed enterocytes at 4 “C prior to homogenization (data not shown) or when lz5I-EGF was ad- ministered intraperitoneally 2 or 10 min before killing (Fig. 2, A and B ) , a similar distribution of the EGF receptor was seen after equilibrium centrifugation, but a greater percentage of radioactivity was associated with the low density vesicles (fractions 3-8).

Localization of the EGF Receptor in Highly Purified Mem- brane Fractions-Because EGF must interact initially with either the BB or LBM surface of the enterocyte, we prepared highly purified membranes by a combination of differential and equilibrium centrifugation to define the localization of EGF binding to extracellular surfaces and intracellular mem- branes. Our conventional separation technique (Ahnen et al., 1982; Cezard et al., 1979) yields a gradient (P3) containing ER-Golgi, intracellular membranes, LBM, and some BB. Treatment with 8 mM CaClz yields a supernatant highly enriched in laterobasal and brush border membranes, but substantially free of ER-Golgi, that can then be separated by equilibrium centrifugation (Nguyen et al., 1987). We now show that this technique not only removes ER-Golgi but other low

3

-LBM- -BB- -ERG -

2 4 6 8 1 0 1 2 1 4 1 6 1 8 2 0

FRACTION NUMBER

25% SORBITOL 60%

FIG. 1. Distribution on sorbitol equilibrium gradient cen- trifugation of EGF binding activity in a crude intestinal mem- brane preparation. The pellet, obtained as described under “EX- perimental Procedures,” was taken up in 6 rnl of binding buffer. 3-rnl aliquots were incubated with ‘“I-EGF (trace) for 90 min or with “‘I- EGF in combination with a 100 M excess of unlabeled EGF. The membranes were then layered over separate 36-ml 25-60% linear sorbitol gradients and centrifuged at 80,000 X g for 16 h to equilib- rium. 2-ml fractions were collected from each gradient and assayed for the associated radioactivity and marker enzymes. Specific EGF binding (M) represents the difference between radioactivity associated with the trace alone and the trace in the presence of excess unlabeled ligand in each of the collected fractions. Location of activ- ities of marker enzymes for membrane fractions is indicated at the top. ERG, ER-Golgi.

density vesicle populations that contain EGF receptors. Of particular importance is that the results of these experiments with purified membranes were consistent with the findings using crude membrane vesicles, even though the binding reaction was performed after the separation of the individual fractions on the equilibrium gradient, rather than prior to cell disruption or membrane separation.

In these studies, the starting material was enterocytes dis- persed from the everted gut sac. When the P3 pellet was placed on top of the gradient and centrifuged to equilibrium, the ER-Golgi (marked by mannosidase 11) and the LBM (by potassium-stimulated phosphatase) could only be partially resolved from one another. As illustrated in Fig. 3, the distri- butions of the ER-Golgi membrane vesicles and LBM vesicles overlapped. Whereas the ER-Golgi activity peaked in fraction 11, and the LBM in fractions 12 and 14, the pattern of EGF binding activity was more complex. EGF binding activity was definitely identified between fractions 4 and 7 and then peaked in fractions 8, 10, and 12. The EGF-EGFR binding correlated with both ER-Golgi and LBM as well as with some membrane vesicles of lighter density. Again, EGF did not appear to bind significantly to brush border membranes.

EGF binding per milligram membrane protein was approx- imately 2 times greater for the PB (LBM/BB) pellet than for the P3 (ER-Golgi/LBM) pellet. Because over 90% of the protein in the PB pellet was associated with brush border membranes, the vesicles capable of binding to EGF may appear to be derived from brush border membranes. Yet, as shown in Fig. 4, this is not the case. When the PB pellet was placed on top of the gradient and centrifuged to equilibrium, the two predominant membrane populations were LBM

1738 Location of Intestinal Epidermal Growth Factor Receptor

2 0 FRACTION NUMBER

25% SORBITOL 60%

E 3

X 2 I E

I I J 0 10 20

FRACTION NUMBER 25% SORBITOL 60%

FIG. 2. Distribution of “‘1-EGF binding after intraperitoneal administration of the ligand. Pellets were prepared from intestinal scrapings from rats that had been injected intraperitoneally with lZ5I-EGF (2 pCi) at 2 ( A ) or 10 ( B ) min prior to killing, as described under “Experimental Procedures.” The crude membrane pellet was taken up in 3 ml of binding buffer and centrifuged as described in the legend to Fig. 1. 2-ml fractions were collected and assayed for the associated radioactivity and marker enzymes. The radioactivity (open squares) was primarily associated with the LBM marker enzyme (closed circles) and lower density vesicles. The peak in fraction 1 represents unbound radiolabeled ligand in the residual cytosol.

40

20

2 , I I - I 0 10 20

FRACTION NUMBER 25% SORBITOL 60%

FIG. 3. EGF binding to purified intracellular and cell sur- face membranes. The P3 pellet obtained as described under “Ex- perimental Procedures” was loaded onto the sorbitol gradients and centrifuged to equilibrium. Specific EGF binding (open squares) and the marker enzymes were measured in each collected fraction after removing the sorbitol. LBM was identified by K+-stimulated phos- phatase (closed circles). These patterns represent typical results ob- tained from a single experiment. Other details are given in the legend to Fig. 1.

(monitored by potassium-stimulated phosphatase, fractions 12-14) and BB (monitored by sucrase, fractions 17 and 18); no peaks of ER-Golgi (mannosidase 11) were seen. The EGF binding activity, except for a small but reproducible lighter density peak in fractions 5-8, coincided closely with the LBM. The specific binding ranged between 4 and 6% in the peak fraction, and, when calculated on the basis of total membrane protein, the LBM (PB) peak from Fig. 4 had a specific binding activity 2-3 times higher than that seen for the peak on the P3 gradient. Indeed, analysis of the data in Fig. 4 comparing the fractional distribution of EGF binding activity and the LBM marker enzyme yielded a correlation coefficient of 0.96 ( p < 0.001), indicating that the vast majority of the active EGF receptor population was associated with the LBM.

In other experiments, crypt and villus tip cells were sequen- tially released by EDTA treatment, harvested, and examined for the ability to bind to lZ5I-EGF at the 55,000 X g membrane

E I>

X r a E E

FRACTION NUMBER 25%- SORBITOL- 60%

FIG. 4. EGF binding activity to LBM-enriched fractions from dispersed enterocytes. The PB pellet devoid of ER-Golgi membranes (see under “Experimental Procedures”) was loaded on top of a 25-60% sorbitol gradient and centrifuged to equilibrium. Other details are given in the legend to Fig. 3. Note the close correlation of EGF binding (open squares) with the LBM (closed circles).

pellet. Crypt specific binding was 8.3%/mg of protein com- pared to 4.0%/mg of protein for villus tip cells. When radio- labeled EGF was injected intraperitoneally 2 or 10 min before killing the rat, a crypt-villus gradient in the distribution of radiolabeled EGF persisted. Most of the binding appeared to correspond with the LBM, when membranes from the crypt pellet were separated on an equilibrium gradient. Finally, a similar preponderance of EGF receptors in the jejunal LBM was seen in Wistar rats, Biobreed Wistar Diabetic (BBW) rats, rats fasted 72 h, and those killed at different circadian phases (data not shown).

Immunohistochemical Localization of the EGF Receptor- The studies of radiolabeled EGF binding strongly suggested that the majority of EGF receptors in the enterocyte is local- ized to the laterobasal surface membranes. However, brush borders may harbor EGF receptors whose binding sites were occupied by endogenous EGF from the lumen and hence unavailable for radioligand binding. To examine this possi-

Location of Intestinal Epider

bility, we localized the receptor in freeze-substituted tissue using well characterized monoclonal and polyclonal receptor antibodies that recognize the rodent EGF receptor. For com- parison, we used rabbit polyclonal antibodies to localize je- junal sucrase, aminooligopeptidase, and epidermal growth factor.

Fig. 5C shows the distribution of sucrase along the villus axis. As expected, this hydrolase localized primarily to the brush border of the mature enterocyte. The major immuno- reactivity appeared above the crypt-villus junction. The lam- ina propria, blood vessels, and nuclei were negative. Similar results were obtained for the hydrolase aminooligopeptidase (not shown). Fig. 5 , A and B, shows the distribution of the EGF receptor along the villus axis. In contrast to sucrase, immunoreactivity determined using the monoclonal antibody (Fig. 5A) was comparable in both crypt and villus cells with the greatest grain density being localized intracellularly in both an infra- and supranuclear position. Brush border stain- ing was negligible. Studies with a rabbit polyclonal antibody against the EGF receptor revealed a similar staining pattern; however, the laterobasal staining was more intense (Fig. 5B), and there was also significant grain distribution over endo- thelial cells lining capillaries and lymphatics in the lamina propria as well as smooth muscle of the muscularis externa. Few grains were observed in the brush border. Although both the monoclonal and polyclonal antibodies specifically recog- nized the EGF receptor in A431 cells as determined by im- munoblot analysis (data not shown), only the polyclonal antibody reacted strongly with the positive control section of liver (not shown). This suggests that the epitope to which the monoclonal antibodv reacts mav not be Dresent in the liver.

, . ..

-mal Growth Factor Receptor 1739

The localization of EGF, the ligand, was confined in the jejunum to the base of the crypts in Paneth cells (Fig. 50) and to occasional villous goblet cells (not shown), confirming recently reported findings (Poulsen et al., 1986).

DISCUSSION

Localization to a Single Type of Surface Membrane in the Enterocyte-We have analyzed the cellular and subcellular distribution of the EGF receptor of mature and immature enterocytes by use of both biochemical and immunocytochem- ical techniques. A striking feature of the enterocyte is that it is polarized, possessing a specialized apical surface facing the intestinal lumen as well as a laterobasal surface exposed to blood and subepithelial cells such as lymphocytes and plasma cells. Since EGF is secreted directly into the intestinal lumen from several different cellular sites of origin, including Brun- ner's glands, Paneth cells, and goblet cells, we expected the majority of EGF surface receptors to be localized to the luminal brush border membranes. Indeed, in a single previous study utilizing partially purified enterocyte brush border prep- arations, the majority of receptors appeared to be localized to the brush border rather than to LBM or microsomal mem- branes (Gallo-Payet and Hugon, 1985)) a finding contrasting with the laterobasal distribution of enteric receptors for vas- oactive intestinal peptide (Dharmsathaphorn et al., 1983) and insulin (Gingerich et al., 1987). Even more recently, specific microvillus receptors for EGF have been reported in fetal and adult but not newborn rats (Thompson, 1988). These recep- tors could be detected at 20 "C but not at 4 "C, and binding was relatively unaffected by pH. However, the membrane preparations used in both studies were not monitored for LBM or microsomal contamination.

By use of highly purified membranes free of significant cross-contamination, we now report the virtual absence of brush border receptors for EGF, based on both binding to purified enterocyte membranes and immunohistochemical studies. Instead, specific enterocyte EGF receptors were iden- tified on the LBM and in intracellular membranes having a slightly lower density than typical ER-Golgi on equilibrium gradients (Figs. 1-3). There was nearly a perfect correlation of the LBM membrane marker with these receptors (Fig. 4). These studies, while not totally excluding the possibility of a minute population of brush border receptors, suggest that the cell surface receptor is restricted or at the very least prefer- entially localized to the LBM. This finding may account for the inconsistent and controversial effects of intragastrically or luminally infused exogenous EGF on intestinal growth and function (Dembinski et al., 1982; Fitzpatrick et al., 1987; Goodlad et al., 1987; Ulshen et al., 1986) and also for why luminal EGF, although absorbed across the intestine in the neonatal rat, does not seem to be absorbed in the adult rat (Olsen et al., 1984; Skov et al., 1984).

Other well differentiated highly polarized cells such as those from eccrine sweat duct epithelium cells (Nanney et al., 1986), jpendymal cells (Nanney et al., 1986), hepatocytes (Burwen et al., 1984; Dunn et al., 1986)) and the Madin-Darby canine kidney cell line (Maratos-Flier et al.. 1987) Dossess EGF

1740 Location of Intestinal Epidermal Growth Factor Receptor

1985). In contrast, hepatocytes and Madin-Darby canine kid- ney cells possess the EGF receptor principally on the latero- basal membrane, a site from which some of the internalized ligand-receptor complexes can cross the cytoplasm to be sub- sequently discharged into the apical lumen (Burwen et al., 1984; Maratos-Flier et al., 1987). These findings are analogous to the intestinal enterocyte since a fraction of systemically administered EGF crosses the gastrointestinal epithelium and appears to be excreted into the lumen (Olsen et al., 1984). The possible mechanism of EGF movement from the intes- tinal lumen to the LBM receptor site of the enterocyte is considered further below.

Intracellular Membrane Location of EGF Receptor (Active and Cryptic)-The enterocytic intracellular low density vesi- cles that harbor the EGF receptor have not been characterized in our studies, but studies in other organ systems may provide insights. By use of the rat isolated perfused liver, Dunn et al. (1986) have also found two major peaks of high affinity EGF binding activity in purified membranes separated in equilib- rium gradients; one peak was associated with the plasma membrane, and the other constituted a population of low density vesicles observed only after exogenous EGF was in- troduced into the system. This suggested that they may rep- resent endosomes. Since the low density intestinal binding sites detected in the present study appeared within minutes after ‘251-EGF (Fig. 2, A and B ) was injected into the perito- neal cavity, they must represent either endosomes or a spe- cialized surface population; yet other cell types containing immunoreactive EGF receptor, such as the endothelial cells within the lamina propria, do not substantially contaminate enterocyte preparations (Weiser et al., 1986), and enterocyte crypt cell plasma membranes migrate to an equilibrium den- sity similar to enterocyte LBM.2 Moreover, because endoge- nous EGF is concentrated in jejunal sites (Paneth (Fig. 5 0 ) and goblet cells), the spontaneous release of this ligand in vivo or during membrane preparation could induce the for- mation of similar receptor-containing endosomes. Indeed, when the circulating and tissue levels of endogenous EGF are barely detectable, such as in rats fasted for 72 h, the total EGF binding activity decreases by 70%, and this low density binding activity disappears (data not shown).

Although the lZ5I-EGF binding experiments (Figs. 1-4) identified the majority of the enterocyte’s active receptor on the LBM, the immunohistochemical analysis revealed a prom- inent specific intracellular receptor pool that appears to pos- sess few active EGF binding sites. Because the antibodies used identify epitopes distinct from the receptor’s ligand binding site, the majority of intracellular receptor species may be devoid of ligand binding sites. The predominant intracel- lular histochemical localization, while not quantitative, sug- gests that there may be even more receptors inside the enter- ocyte than on its surface (Fig. 5, A and B ) . A similar discrep- ancy between the pools of active and immunoreactive receptor has been recently reported in the A431 epidermoid carcinoma cell line where as many as 8045% of EGF receptors were associated with several different intracellular organelles and less than 15% of the total immunologically reactive EGF receptors were capable of binding the EGF ligand (Carpentier et al., 1986). Cryptic receptors have also been observed in OC15 embryonal carcinoma cells (Weller et al., 1987).

Possible Modes of Luminal to LBM Transport of EGF- Because EGF is secreted from Brunner’s duodenal glands and from crypt Paneth cells into the intestinal lumen, it seemed logical that the EGFR would be localized at the luminal surface where EGF binding could regulate replication and

L. A. Scheving and G. M. Gray, unpublished observations.

enterocyte function. In light of our results, the physiological significance of EGF within the intestinal lumen must be reexamined. Luminally infused exogenous EGF has been dem- onstrated in some but not in all cases to modulate intestinal growth and function. In one instance, intraintestinal infused EGF stimulated ornithine decarboxylase activity and DNA synthesis not only in the surgically isolated loop of ileum receiving infusate but also in jejunum that was denied access to the ligand from the luminal surface (Ulshen et al., 1986). This suggests that the effects of intraluminal EGF may be neuroendocrine in nature or involve its transport from the lumen across the enterocyte where is could have access to both the LBM receptor sites and to the extracellular com- partment. Possibly, receptors are recruited or transported to the lumenal surface under special physiological conditions, such as in the intestinal adaptation subsequent to surgical resection. Such an adaptive mechanism could involve other cell types (Paneth, tuft, neuroendocrine, or membranous ep- ithelial (M) cells) which, being much less numerous than enterocytes, are difficult to analyze in the intact intestine and cannot be readily isolated.

Whatever the exact mechanism whereby luminal EGF in- teracts with the intestine, our experiments do bring into focus an alternative hypothesis: intraluminal EGF may gain access to its receptor at the LBM surface by virtue of local move- ment. Under normal conditions, the tight junctions effectively seal the lateral surfaces of adjacent enterocytes, restricting the movement of molecules and particles between luminal and perilateral spaces. However, there is evidence that the luminal-laterobasal boundary is not inviolable and that lu- minal contents may gain access to the LBM. Recent electron microscopic studies of normal intestine indicate that under special circumstances small gaps can arise in either crypt or villus cells, thereby exposing luminal contents to lateral sur- face membranes. In crypt cells, these gaps arise because tight junctions disassemble and re-form during cell division (Mar- cia1 and Madara, 1987; Tice et al., 1979). Acetylcholine trig- gering of crypt goblet cell secretion produces a similar break- down in goblet cell tight junctions, creating luminal-pericel- lular communications (Marcia1 and Madara, 1987). Since most of the intestinal EGF is normally secreted by Brunner’s glands into ducts that empty into the duodenal crypts or by Paneth and goblet cells, these intercellular gaps may provide a conduit for luminal EGF to reach its LBM receptor. In villus cells, gaps arise when enterocytes at the villus tips are shed into the intestinal lumen. Pathological conditions caus- ing erosions or ulcerations and a loss of integrity of the tightly connected blanket of enterocytes may assure the interaction of lateral membranes of crypt or villus cells with luminal molecules (Olsen et al., 1984; Skov et al., 1984). Thus, the seemingly paradoxical distribution of receptor and ligand in vivo may reflect the existence of a regulatory barrier between luminal EGF and its LBM surface receptor. Definition of the mechanisms whereby this barrier is modulated may be crucial for our understanding of the maintenance of physiological and pathological enterocyte function.

REFERENCES

Ahnen, D. J., Santiago, N. A., Cezard, J.-P., and Gray, G. M. (1982)

Alquist, J. (1972) Acta Pathol. Microbial. Scand. Sect. A . Pathol. 80,

Bunven, S. J., Barker, M. E., Goldman, I. S., Hradek, G. T., Raper,

Carpenter, G., and Cohen, S. (1979) Annu. Reu. Biochem. 48, 193-

Carpentier, J.-L., Rees, A. R., Gregoriou, M., Kris, R., Schlessinger,

J. Biol. Chem. 257, 12129-12135

177-184

S. E., and Jones, A. L. (1984) J. Cell Biol. 99, 1259-1265

216

J., and Orci, L. (1986) Exp. Cell Res. 161, 312-326

Location of Intestinal Epidermal Growth Factor Receptor 1741

Cezard, J.-P., Conklin, K. A., Das, B. C., and Gray, G. M. (1979) J. Nanney, L. B., Stoscheck, C. M., and King, L. E., Jr. (1986) J. Cell Biol. Chem. 254,8969-8975 Biol. 103 , 155a

Chabot, J. G., Guthrie, P. D., and Hugon, J. S. (1983) Comp. Biochem. Nguyen, T. D., Broyart, J.-P., Ngu, K. T., Illescas, A., Mircheff, A. Physiol. 7 4 , 247-252 K., and Gray, G. M. (1987) J. Membr. Biol. 9 8 , 197-205

Danscher, G., and Norgaard, J. 0. R. (1983) J. Histochem. Cytochem. Olsen, P. S., Poulsen, S. S., Kirkegaard, P., and Nexo, E. (1984) 31,1394-1398 Gastroenterology 87, 103-108

Dembinski, A., Gregory, H., Konturek, S. J., and Polanski, M. (1982) Poulsen, s. s., Nexo, E., Skov, p., Olsen, s. S., Hess, J., and K i r k - J. Physiol. 325, 35-42 gaard, P. (1986) Histochemistry 85, 389-394

Dharmsathaphorn, K., Harms, V., Yamashiro, D. J., Hughes, R. J., Rao, c. v., Ramani, N., Chegini, N., Stadig, B. K., Carman, F. R., Binder, H. J., and Wright, E. M. (1983) J. C h . Invest. 7 1 , 27-35 Jr., Woost, p. G., Schultz, G. S., and Cook, C. L. (1985) J. Biol.

Dunn, W. A., Connolly, T. P., and Hubbard, A. L. (1986) J. Cell Biol. 260, 1705-1710 102,24-36 Scatchard, G. (1949) Ann. N. Y. Acad. Sci. 51,660-672

Fitzpatrick, L. R., Wang, P., and Johnson, L. R. (1987)Am. J. Physiol. Scheving, L. A., Yeh, y. c., Tsai, T. H., and Scheving, L. E. (1980) 252, G209-G214 Endocrinology 106 , 1498-1503

Forgue-Lafitte, M. E., Laburthe, M., Chamblier, M. C., Moody, A. J., Scheving, L. A.3 Scheving, L. E., Tsai, T. H., and Houk w. s. (1987)

Gallo-Payet, N., and Hugon, J, S. (1985) Endocrinology 116 , 194- Skov, p.9 Olsen, s. s., Poulsen, p.9 Kirkegaard, p.9 and NexO, E.

Gingerich, R, L., Gilbert, W. R,, Cornens, p, G,, and Gavin, J. R., 111 Sternberger, L. A. (1979) zmmunocytochemist~ 2nd Ed., P. l o 4 9 John

and Rosselin, G. (1980) FEBS Lett. 114, 243-246 Peptides 8, 347-353

201 (1984) Dig. Dis. Sci. 29, 615

(1987) Diabetes 36, 1124-1129

K. G., and Wright, N. A. (1987) Gut 28,37-44

and Pearse, A. G. E. (1978) Gut 19, 408-413

Histochem. Cytochem. 3 1 , 938-944

(1987) J. Cell Biol. 105 , 1595-1601

Wiley & Sons, New York Goodlad, R. A., Wilson, T. J. G., Lenton, W., Gregory, H., McCullagh, ~ ~ m ~ ~ ~ : , J d ~ ~ ~ ~ ~ ~ ~ ~ l h ~ ~ ~ ~ 2 ~ 4 ; ~ ~ ~ ~ ~ ~ ~ ~ cell 1,

Heitz, p. u.3 Kasper, M.3 Van Noorden, s., Polak, J. M., Gregory, H., ~ ~ l ~ i ~ ~ i , D. R. p., ~ ~ b b ~ ~ d , s, c, , Robbins, p, w., and Touster, 0. 293-316

(1982) J. Biol. Chem. 267 , 3660-3668

terology 9 1 , 1134-1140

Cytol. 101 , 1-49

Hokate, c. s., Jackson, p., Cowen, p. N.7 and Bird, c. c. (1983) J. Ulshen, M. H., Lyn-Cook, L. E., and &as&, R. H. (1986) Gastroen-

Maratos-Flier, E., Yang-Kao, c. y.2 Verdin, E. M.9 and King, G. L. Weiser, M. M., Walters, J. R. F., and Wilson, J. R. (1986) Znt. Rev.

Marcia4 M. A., and Madam, J. L. (1987) Lab. h e s t . 56,424-431 Weller, A., Meek, J., and Adamson, E. D. (1987) Development 100 , Michael, N. L., Rothbard, J. B., Shiurba, R. A., Linke, H. K., 351-363

Schoolnik, G. K., and Clayton, D. A. (1984) EMBO J. 3, 3165- Yeh, Y. C., Scheving, L. A., Tsai, T. H., and Scheving, L. E. (1981) 3 168 Endocrinology 109, 644-652