Epithelial membrane protein-1, peripheral myelin protein 22, and lens membrane protein 20 define a novel gene family.

Peripheral myelin protein 22 (PMP22) is expressed in many tissues but mainly by Schwann cells as a component of compact myelin of the peripheral nervous system (PNS). Mutations affecting PMP22 are associated with hereditary motor and sensory neuropathies. Although these phenotypes are restricted to the PNS, PMP22 is thought to play a dual role in myelin formation and in cell proliferation. We describe the cloning and characterization of epithelial membrane protein-1 (EMP-1), a putative four-transmembrane protein of 160 amino acids with 40% amino acid identity to PMP22. EMP-1 and PMP22 are co-expressed in most tissues but with differences in relative expression levels. EMP-1 is most prominently found in the gastrointestinal tract, skin, lung, and brain but not in liver. In the corpus gastricum, EMP-1 protein can be detected in epithelial cells of the gastric pit and isthmus of the gastric gland in a pattern consistent with plasma membrane association. EMP-1 and PMP22 mRNA levels are inversely regulated in the degenerating rat sciatic nerve after injury and by growth arrest in NIH 3T3 fibroblasts. The discovery of EMP-1 as the second member of a novel gene family led to the identification of the lens-specific membrane protein 20 (MP20) as a third but distant relative. The proteins of this family are likely to serve similar functions possibly related to cell proliferation and differentiation in a variety of cell types.

The 22-kDa peripheral myelin protein (PMP22) 1 is a hydrophobic protein of 160 amino acids with four predicted transmembrane domains (1,2). Point mutations in PMP22 and aberrant expression of the PMP22 gene are associated with various hereditary peripheral motor and sensory neuropathies (3). In particular, the spontaneous mouse mutants Trembler and Trembler-J carry point mutations in the pmp22 gene (4,5). In humans, the majority of patients suffering from the autosomal dominant demyelinating neuropathy Charcot-Marie-Tooth disease type 1A (6) bear a 1.5-megabase intrachromosomal duplication of chromosome 17p11.2-12 that includes the PMP22 gene (7)(8)(9)(10). Rare point mutations in the PMP22 gene have also been found in non-duplication Charcot-Marie-Tooth disease type 1A patients (11)(12)(13) and in the severe congenital peripheral neuropathy Dejerine-Sottas syndrome (14). Furthermore, the relatively mild, recurrent peripheral neuropathy with liability to pressure palsies is associated with the reciprocal deletion to the Charcot-Marie-Tooth disease type 1A duplication (15,16).
As anticipated from the phenotype of PMP22 mutant organisms, PMP22 is mainly expressed by myelinating Schwann cells in the peripheral nervous system (PNS) where it is incorporated into compact myelin (1,(17)(18)(19). PMP22 mRNA and protein have also been found in motorneurons, and transcripts have been identified in various adult tissues, including the brain, intestine, lung, and heart (8, 20 -22). Furthermore, PMP22 mRNA expression is widespread during mouse embryonic development (23). Tissue-specific expression and regulation of PMP22 is controlled by a complex genetic mechanism involving two alternative promoter sequences (24,25). While the distal promoter is specifically activated in myelinating Schwann cells, the more proximal promoter was found to be active in all known PMP22-expressing tissues (24).
Tissue culture experiments using NIH 3T3 cells and primary dermal fibroblasts revealed that PMP22 is up-regulated under growth arrest conditions, e.g. serum deprivation or density growth arrest, suggesting a potential role for PMP22 in cell proliferation (20,27). 2 In support of this hypothesis, recent studies employing retrovirus-mediated gene transfer of PMP22 into cultured Schwann cells suggest a pronounced influence of PMP22 expression on the length of the G 1 phase (28).
Based on these findings, it was proposed that PMP22 serves a general role in cell physiology and an additional specialized function in PNS myelin (29). However, all known mutations affecting the PMP22 are associated with a phenotype restricted to the PNS, and no consistent abnormalities in non-neural tissues have been detected, even in genetically engineered mice that are completely devoid of PMP22 (30). The most likely explanation for these apparently contradictory results is to postulate specific mechanisms that can compensate for the lack of PMP22 in non-neural tissues. Such processes may involve molecules that are structurally and/or functionally related to PMP22. However, although the putative membrane topology of PMP22 is similar to the gap junction-forming connexin protein family (31) or the tight junction component occludin (32), PMP22 does not display convincing amino acid sequence identity with any other known protein.
In this study, we report the identification of a PMP22-related transcript and the characterization of its encoded protein, which we have designated epithelial membrane protein-1 (EMP-1) based on its tissue expression pattern. EMP-1 and PMP22 are significantly related in their overall structure and primary amino acid sequences and define a novel gene family that also includes a more distant relative, the lens-specific membrane protein MP20 (33).

EXPERIMENTAL PROCEDURES
Animals and Surgery-Male SIV rats (8 weeks old; University of Zurich, Switzerland) were anesthetized by interperitoneal injection of a mixture of ketamine and choral hydrate (17). Both sciatic nerves were exposed, and the right nerve was cut approximately 2 mm distal to the hip joint (17). Animals were sacrificed 4 days after injury in a CO 2 atmosphere. For isolation of RNA and protein, tissues were excised and either lysed directly into 5 M GT buffer (5 M guanidinium isothiocyanate (Ultrapure, Life Technologies, Inc.), 25 mM sodium citrate, and 0.5% sodium N-lauryolsarcosine), or 8 M urea, or snap-frozen in liquid nitrogen.
Cloning and Sequencing of EMP-1 cDNA-Recombinant clones were isolated from a fetal rat intestine cDNA library. Plasmid inserts were analyzed by automated sequencing in both orientations. The resulting nucleotide sequences were compared with the GenBank TM data base to identify known and related sequences. EMP-1 cDNA was identified as being significantly related to rat PMP22, and the 981-base pair cDNA clone was sequenced on both strands. The sequence was screened for open reading frames by translation in all frames. The longest open reading frame was compared with protein data bases and again showed significant similarity to PMP22. Computer-assisted sequence analysis was performed using the GCG software package to produce hydrophobicity profiles, identify signal peptide cleavage sites, and predict surface probability. The EMP-1 and PMP22 polypeptide sequences were compared using the Pileup program and depicted graphically by Prettyplot to show amino acid residues conserved between EMP-1 and the PMP22 species homologues.
In Vitro Transcription and Translation-Both the EMP-1 and mouse PMP22 cDNAs were cloned into the pcDNA-1 expression vector downstream of the bacteriophage T7 promoter region and used for in vitro transcription/translation assays. mRNA was produced from 0.5 g of DNA with T7 polymerase and translated in vitro in the presence or absence of canine microsomal membranes (CMM) using a reticulocyte lysate system (TNT, Promega). The translation products were labeled metabolically by including [ 35 S]methionine in the reaction. The specificity of the reaction was tested using the vector pcDNA-1 as a control. The efficiency of the CMM was tested with the control reagents ␣-factor (glycosylation) and prolactin (signal peptide cleavage) according to the manufacturer's instructions (TNT, Promega). One-tenth of the translation product was denatured in 1% SDS and incubated for 4 h at 37°C with 1 unit of N-glycosidase F in 50 mM sodium phosphate (pH 7.2), 12.5 mM EDTA, 2.5 mM sodium azide, 25% glycerol, and 0.2% SDS. The proteins were separated by reducing 15% SDS-PAGE, and the gels were subsequently fixed 30 min in 50% methanol, 10% acetic acid and treated with enhancer (NEF-981G, Dupont). After drying the gels were exposed to x-ray film (RX, Fuji) overnight at Ϫ70°C.
Isolation of Total RNA and Northern Blot Analysis-Total RNA was extracted from rat tissues using a modified acid phenol method. Briefly, tissues were homogenized into GT buffer (34). The lysate was cleared and extracted twice with phenol/chloroform (1:1). The RNA was precipitated, resuspended in diethyl pyrocarbonate-treated H 2 O, and quantified by A 260 /A 280 measurement. Ten g of total RNA was loaded onto denaturing 1.2% agarose formaldehyde gels. Separated RNA was transferred to nylon membrane (Hybond N, Amersham Corp.) by capillarity and cross-linked with 240 mJ of UV irradiation in a Stratalinker (Stratagene). Equal loading and transfer to the membrane was assessed by ethidium bromide staining. Membranes were prehybridized for 6 h and hybridized 36 h at 42°C in a solution containing 50% formamide. cDNA fragments of PMP22 and EMP-1 containing the entire open reading frames were labeled with 32 P-dCTP by random hexamer priming (Oligolabeling kit, Pharmacia Biotech Inc.). Northern blots were washed at high stringency and exposed to x-ray film (RX, Fuji) for 12-72 h.
Tissue Culture-Rat Schwann cells were isolated from the sciatic nerve of neonatal rats using the method of Brockes et al. (35) with modifications as described previously (19). The mitogen-expanded primary rat Schwann cells were cultured on poly-L-lysine coated culture plates in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum, 20 g/ml crude glial growth factor, 5 g/ml forskolin, and 50 g/ml gentamicin (19). The Schwann cell-derived cell line, D6P2T (36), was grown in plastic culture dishes in DMEM containing 5% fetal calf serum and 50 g/ml gentamicin. The cells were split into two groups and cultured in the same medium without forskolin. After 3 days, the cells had reached 70% confluency; one group of cells were treated with 20 g/ml forskolin for 36 h, while the second was maintained in the absence of forskolin. Subsequently, the cells were harvested in 5 M GT buffer, and total RNA was extracted as described above.
NIH 3T3 fibroblasts were cultured, and growth was arrested by serum deprivation as described previously (24). Exponentially growing and growth-arrested cells were harvested into 5 M GT buffer.
Preparation of the EMP-1 Expression Construct and Transient Transfection of COS Cells-The EMP-1 cDNA shown in Fig. 1 was subcloned into the EcoRV site of the pcDNA1 (InVitrogen) expression vector, downstream of the cytomegalovirus promoter. The parent vector without cDNA insert was used in the negative control transfections. Recombinant DNA was purified by QIAGEN column isolation and quantitated by A 260 . COS cells exponentially growing in DMEM containing 10% fetal calf serum were trypsinized, washed in phosphatebuffered saline, and pelleted at 800 g. 1.5 ϫ 10 6 cells were resuspended in 200 l of phosphate-buffered saline containing 5 g of vector DNA. The cells were chilled on ice for 5 min, transferred to a 4-mm gap electroporation cuvette (Bio-Rad), and electroporated with 300 V and 125 microfarads. The transfected cells were chilled for 5 min on ice and split into seven 35-mm culture dishes containing 2 ml of DMEM and 10% fetal calf serum and cultured for 48 h. The cells were then washed with Tris-buffered saline, fixed in DMEM containing 2% paraformaldehyde for 30 min, washed with Tris-buffered saline, and permeabilized for 30 min in Tris-buffered saline containing 0.1% saponin. Unspecific binding sites were blocked for 30 min at room temperature in Trisbuffered saline, 2% bovine serum albumin, 0.1% porcine skin gelatin (type A Sigma), 2% goat serum, and 0.1% saponin. The cells were incubated with the anti-EMP-1 antibodies diluted 1:500 in blocking buffer containing 0.02% saponin overnight at 4°C followed by fluorescein isothiocyanate-labeled goat anti-rabbit immunoglobulin (Cappel) at 1:200 in blocking buffer with 0.02% saponin. After washing with Tris-buffered saline, coverslips were mounted using AF1 (Citifluor), and immunoreactivity was visualized by confocal microscopy with a Bio-Rad MRC-600 scanner in conjunction with a Zeiss Axiophot fluorescence microscope using Imaris image processing software (Bitplane AG, Technopark Zü rich, Switzerland). Preimmune serum (1:500) as primary antiserum was used as a negative control.
Amino acids are numbered according to the cDNA-predicted polypeptide shown in Fig. 1. A C-terminal cysteine residue was added to the loop 1 peptide for coupling purposes. The peptides were coupled to keyhole limpet hemocyanin as described previously (17). The conjugates were used to immunize New Zealand White rabbits with Freund's complete adjuvant, and the animals were boosted 4 times with 500 g of peptide and incomplete adjuvant at 2-week intervals. Blood was taken from the animals, and serum was isolated. The activity of the immune serum was tested on the immunogen by solid phase ELISA.
Protein Preparation and Western Blot Analysis-Rat tissues were homogenized into 8 M urea and cleared at 10,000 ϫ g for 10 min at 4°C. The protein concentration of the supernatant was assessed by Bradford assay, and 50 g of protein were denatured by heating to 95°C for 3 min in sample buffer containing 2% ␤-mercaptoethanol and loaded on a 12% SDS-PAGE. The proteins were electrotransferred to nitrocellulose membrane (Schleicher & Schuell) using a semi-dry blotter (Bio-Rad). Membranes were stained with Ponceau S to test transfer efficiency. Blots were blocked with 0.15% casein in phosphate-buffered saline containing 0.2% Tween 20 (Sigma), incubated with the polyclonal rabbit sera at a dilution of 1:500 in blocking buffer followed by a 1-h incubation with horseradish peroxidase-labeled goat anti-rabbit immunoglobulin (1:5000, Sigma). Detection was by chemiluminescence (ECL, Amersham) and exposure of x-ray film (RX, Fuji).
Tissue Preparation and Immunofluorescence-Stomach was removed from adult SIV rats, cut into blocks 10 by 5 mm, and snap-frozen into OCT mounting medium (Tissue Tech) in isopentane on liquid nitrogen. Ten-m frozen sections were cut and thaw-mounted onto slides coated with 0.5% gelatin and 0.05% potassium chromate. The sections were fixed in 1% paraformaldehyde for 5 min and washed in phosphatebuffered saline. Unspecific binding sites were blocked for 1 h with goat serum diluted 1:50 in phosphate-buffered saline. Sections were incubated overnight at 4°C or for 4 h at room temperature with a 1:500 dilution of primary antiserum in phosphate-buffered saline containing 0.2% Tween 20 followed by 2 h with a 1:200 dilution of Texas Redlabeled donkey anti-rabbit immunoglobulin (Jackson Immunoresearch Laboratories). After extensive washing in phosphate-buffered saline containing 0.2% Tween 20, sections were coated with AF1 (Citifluor), and coverslips were mounted. Immune reactivity was visualized by high resolution confocal microscopy. Preimmune serum (1:500) as primary antiserum was used as a negative control.

RESULTS
Cloning of an EMP-1 cDNA and Predicted Structure of the EMP-1 Protein-During a fetal rat intestine cDNA sequencing project, a 981-base pair cDNA containing an open reading frame of 480 nucleotides was identified (Fig. 1). The predicted EMP-1 polypeptide of 160 amino acid residues has a calculated molecular weight of approximately 18 kDa, and computer-assisted analysis reveals that amino acid residues 1-28, 64 -89, 95-117, and 134 -157 represent four hydrophobic, potentially membrane-spanning domains (Fig. 2B). The amino-terminal 16 amino acids have the characteristics of a signal peptide including a signal peptidase cleavage site after alanine 16 ( Fig. 1). Furthermore, a single consensus sequence for N-linked glycosylation is present at asparagine 43 (Figs. 1 and 2B).
EMP-1 Is a Member of a Gene Family-Comparison of the EMP-1 cDNA sequence to the GenBank TM data base identified PMP22 as the closest known relative with a nucleotide identity of 58% over the open reading frame (data not shown). Both the predicted EMP-1 protein and PMP22 are polypeptides of 160 amino acids that show 40% amino acid identity ( Fig. 2A). The putative four membrane-spanning regions of EMP-1 and PMP22 are particularly well conserved. The first and second of these hydrophobic domains exhibit the highest degree of amino acid identity at 54 and 67%, respectively, while the third and fourth are only 30 and 37% identical. Fig. 2B depicts a theoretical model of the EMP-1 protein structure based on the hydrophobicity profile and the suggested structure of PMP22. Filled circles represent identical amino acid residues that are shared by rat EMP-1 and rat PMP22, while divergent residues are shown as open circles. The positions of the amino acids in PMP22 known to cause hereditary motor and sensory neuropathies (21) when mutated are highlighted in the EMP-1 sequence as diamonds (Fig. 2B). Interestingly, all of these mutations lie within the putative membrane-spanning domains, and five of the six residues are conserved in rat EMP-1. The conservation of these amino acid residues suggests that they may be of functional significance.
Additional data base searches with the EMP-1 and PMP22 sequences revealed that both display 30% amino acid identity to the lens fiber cell protein MP20 (33). MP20 is a 173-amino acid protein with similar structural features to EMP-1 and PMP22 (Fig. 2C). If MP20 is compared with PMP22 and EMP-1 simultaneously, the amino acid identity increases to 36% including strongly conserved motifs in the putative transmembrane domains (Fig. 2C).
EMP-1 Is Glycosylated in Vitro but Its Signal Sequence Is Not Cleaved-The EMP-1 cDNA shown in Fig. 1 was transcribed and translated in vitro (Fig. 3). The translation product in the absence of CMMs has an apparent molecular mass of approximately 18 kDa, which is in agreement with the calculated molecular mass of the EMP-1 protein. Translation in the presence of CMM results in a 4 -6-kDa increase in molecular mass consistent with the presence of the single putative N-linked glycosylation site in the EMP-1 amino acid sequence (Fig. 3). This migrational shift can be reversed by deglycosylating the translation product with N-glycosidase F (Fig. 3). The deglycosylated protein migrates identically to the unglycosylated EMP-1 protein, suggesting that the putative N-terminal signal peptide is not removed during EMP-1 protein biosynthesis. A similar modification of PMP22 was seen with CMM, confirming previous reports of endogenous PMP22 carrying an uncleaved signal peptide (37).
Expression Patterns of EMP-1 mRNA and PMP22 mRNA Are Qualitatively Similar but Quantitatively Different-To elucidate the distribution of EMP-1 mRNA in the rat, we used the EMP-1 cDNA (Fig. 1) to probe Northern blots of total RNA extracted from various tissues. EMP-1 transcripts can be found in all organs examined with the exception of the liver (Fig. 4, a  and b). The most prominent EMP-1 mRNA expression is observed in tail-derived skin and in the gastrointestinal tract. To examine a potential specific regional expression pattern, RNA was extracted from different regions of the gastrointestinal tract. Interestingly, we found that EMP-1 mRNA is not uniformly expressed throughout the stomach of the rat. The fundic region exhibits high levels of EMP-1 mRNA, while expression in the corpus and pylorus is much lower. In the intestinal tract, the cecum and large intestine (colon and rectum) contain the highest levels of EMP-1 mRNA. EMP-1 transcripts are also detectable throughout the small intestine but at far lower levels than in the fundus of the stomach, the cecum, and large intestine. Considerable amounts of EMP-1 mRNA, similar to the expression in the duodenum, are also found in the brain and lung. Low level EMP-1 expression is detectable in the heart, kidney, spleen, thymus, and skeletal muscle.
All tissues expressing EMP-1 mRNA contain 2.8-kb EMP-1 transcripts. In some regions of the gastrointestinal tract, however, including the fundus, ileum, cecum, and colon, additional transcripts of approximately 1.7 kb hybridize with the EMP-1 cDNA (Fig. 4a). Prolonged washing of the blots at high stringency did not result in the preferential loss of one of the signals relative to the other (data not shown), hence, we favor the interpretation that both the 2.8-and 1.7-kb transcripts are derived from the EMP-1 gene. Further studies of the differently sized transcripts will determine if they result from the use of alternative polyadenylation sites or arise by alternative splicing.
The EMP-1-probed Northern blot was stripped and reprobed with labeled rat PMP22 cDNA (Fig. 4, c and d). The results show that the tissue distribution of PMP22 mRNA and EMP-1 mRNA is similar but that there are subtle differences in their relative expression levels. PMP22 and EMP-1 transcripts are co-expressed to high levels in the skin, fundus of the stomach, cecum, colon, rectum, and duodenum. However, while the EMP-1 mRNA level is relatively high in colon compared with the rectum, PMP22 mRNA is low. Furthermore, PMP22 mRNA is more prominently expressed in the lung than EMP-1 mRNA but is relatively underrepresented in the brain (Fig. 4). Neither EMP-1 or PMP22 transcripts could be detected by Northern analysis of liver RNA.
EMP-1 mRNA and PMP22 mRNA Are Differently Regulated after Sciatic Nerve Injury-The highest levels of PMP22 are found in myelinating Schwann cells of the PNS, and expression is down-regulated in the distal portion of the rat sciatic nerve following crush or cut injury (1,2,17,38). We have examined EMP-1 mRNA levels in the sciatic nerve under the same conditions. The sciatic nerves of 2-month-old rats were bilaterally exposed and cut unilaterally. Four days after injury, the degenerating distal portion of the traumatized nerve and, as a control, the undamaged contralateral nerve were removed, and total RNA was isolated. Northern blot analysis reveals that EMP-1 transcripts are present at considerably lower levels in the adult sciatic nerve than PMP22 transcripts; Fig. 5 represents exposures of 1 h for PMP22 and approximately 36 h for EMP-1. Furthermore, EMP-1 expression increases in the distal nerve following injury, in sharp contrast to PMP22 (Fig. 5). These findings demonstrate that, although EMP-1 and PMP22 are coexpressed in PNS nerves, they appear to be differently regulated.
EMP-1 and PMP22 mRNA Expression in Schwann Cells Is Inversely Regulated in Vitro-In order to substantiate the hy- FIG. 3. In vitro transcription and translation of EMP-1 and PMP22 cDNAs. The [ 35 S]methionine metabolically labeled proteins were separated by reducing 15% SDS-PAGE. pcDNA-1 was used as a control, and no nonspecific proteins can be detected. The EMP-1 cDNA generated an 18-kDa protein that shows a tendency to aggregate (EMP-1). Transcription and translation of the EMP-1 cDNA in the presence of CMM results in a reduced rate of migration of the protein (EMP-1 ϩ CMM). This reduced migration is reversed by deglycosylation with N-glycosidase F (EMP-1 deglycosylated). The PMP22 cDNA generates an 18-kDa protein whose apparent molecular weight increases by 4 -6 kDa when the reaction is performed in the presence of CMM (PMP22 ϩ CMM). Treatment with N-glycosidase F reduced the molecular mass back to 18 kDa (PMP22 deglycosylated). Neither EMP-1 nor PMP22 are substrates for signal peptidase in vitro as indicated by the identical migration rate of the unglycosylated (EMP-1, PMP22) and deglycosylated translation products (EMP-1 deglycosylated, PMP22 deglycosylated). Prolactin (a) is a substrate for signal peptidase; approximately 50% of the protein is processed when translated in the presence of CMM (c) (Promega technical manual). ␣-factor was used as a control for N-linked glycosylation and was completely modified in the reactions (b) (Promega technical manual).

FIG. 4. Tissue distribution of EMP-1 and PMP22 mRNAs in the rat.
Northern blot analysis with a radiolabeled EMP-1 probe shows high expression of 2.8-kb transcripts in the cecum, colon, rectum, fundus, and ileum (a). Lower levels of expression are observed in the duodenum and jejunum of the small intestine and the corpus and pylorus of the stomach (a). Additional transcripts of 1.7 kb are found in the fundus, ileum, cecum, and colon (a). In extraintestinal tissues, EMP-1 mRNA levels are high in the skin, whereas in the brain and lung, expression is comparable with the duodenum (b; panels a and c are 20-h exposures, and panels b and d are 48-h exposures of the same blot). PMP22 mRNA is also highly expressed in the intestine (c); its 1.8-kb transcript is most prominent in the rectum and cecum, where expression is comparable with that of PMP22 in the lung (c, d). Ten g of total RNA was loaded per lane, and equal loading was verified by ethidium bromide staining (data not shown).
pothesis of inverse regulation of EMP-1 and PMP22, we compared EMP-1 and PMP22 mRNA levels in mitogen-expanded primary rat Schwann cells (pSC) and D6P2T Schwann cells grown in the presence or absence of forskolin. Forskolin has been shown to induce the expression of PMP22 and other myelin proteins in cultured Schwann cells via a mechanism proposed to partially mimic axon-Schwann cell interactions that occur during myelination (2,19,39). Total RNA was isolated from mitogen-expanded pSC and the D6P2T cell line, grown in the presence or absence of forskolin. As expected, PMP22 mRNA is up-regulated by forskolin in pSC and D6P2T cells (Fig. 5B). In contrast, EMP-1 mRNA levels are decreased under the same conditions. This regulation is particularly prominent in the D6P2T cell line where EMP-1 mRNA is reduced in the presence of forskolin to barely detectable levels.
EMP-1 mRNA and PMP22 mRNA Are Inversely Regulated by Growth Arrest in NIH 3T3 Cells-PMP22 was originally identified as being identical to the growth arrest-specific mRNA gas-3 in NIH 3T3 fibroblasts (27). We confirmed this finding by showing that upon serum starvation and subsequent growth arrest, NIH 3T3 fibroblasts exhibit increased PMP22 mRNA expression (Fig. 5C) (20). Again, in contrast to the regulation of PMP22, EMP-1 mRNA levels are strongly decreased under identical experimental conditions (Fig. 5C).
The 25-kDa EMP-1 Protein Is Highly Expressed in Rat Intestine-In order to determine the tissue distribution of the EMP-1 protein, we have raised separate polyclonal anti-peptide antisera against the first (loop 1) and second (loop 2) putative extracellular domains of the predicted EMP-1 polypeptide (Fig. 2B). Both antisera recognize a protein of approximately 25 kDa by Western blotting of various gastrointestinal tract tissue lysates (Fig. 6) but the anti-loop 1 antibodies are considerably more efficient. The labeling of immunoreactive proteins on Western blots can be blocked by preincubation of the antiserum in the presence of 250 g/ml of immunogen peptide, confirming specificity of the signal (Fig.  6A). Blocking is specific for the immunogen and is not effected by preincubation with the same concentration of a different peptide (Fig. 6A). The calculated molecular mass of the core EMP-1 protein is approximately 18 kDa, which can be confirmed by in vitro transcription and translation of the cDNA (Fig. 3). The presence of a putative N-linked carbohydrate chain conceivably results in a protein with an apparent molecular mass of 25 kDa on reducing SDS-PAGE. However, the possibility of additional post-translational modifications of the EMP-1 protein cannot be excluded.
The most prominent expression of EMP-1 protein is seen in the stomach, with lower levels being detectable in the cecum and large intestine. Expression in the duodenum and jejunum

FIG. 5. Regulation of EMP-1 and PMP22 mRNA expression by sciatic nerve injury and in cultured cells.
A, Northern blot analysis of EMP-1 mRNA reveals an increased expression in the degenerating distal part of the injured sciatic nerve (4 days after nerve cut) compared with normal control nerve. PMP22 expression is considerably higher than that of EMP-1 in the normal nerve (1-h exposure using the PMP22 probe compared with 36 h for the EMP-1 probe). In contrast to EMP-1, PMP22 mRNA is dramatically reduced in the distal nerve after injury. B, cultured, mitogen-expanded primary rat Schwann cells (pSC) and D6P2T Schwann cells display reduced EMP-1 expression following forskolin treatment. In contrast, PMP22 mRNA expression is increased under the same conditions. C, serum starvation-induced growth arrest of NIH 3T3 cells results in reduced EMP-1 mRNA expression and an increase in PMP22 expression relative to exponentially growing cells. Northern blot analyses were performed on the same blot (10 g of total RNA/sample), which was stripped between hybridizations.

FIG. 6. Expression of EMP-1 protein in the rat intestine.
A, two rabbit anti-EMP-1 peptide antisera raised against each of the putative extracellular loops 1 and 2 of EMP-1 recognize a 25-kDa protein in the corpus gastricum (50 g of protein lysate analyzed by 12% SDS-PAGE and Western blotting). The immune reactivity of the anti-loop 1 antiserum was blocked by preincubation with 250 g/ml of the immunogen but not by the loop 2 peptide. B, strong expression of EMP-1 protein is found in the stomach and large intestine, and lower levels are present in the lung. Detection of a signal in the small intestine requires prolonged reaction time of the enzymatic detection system (data not shown).
of the small intestine is considerably lower than in the other regions of the intestinal tract, in accordance with the reduced mRNA levels found in these tissues (Figs. 4 and 6A). Very low levels of the 25-kDa EMP-1 protein can also be detected in the lung (Fig. 6A), spleen, and thymus (data not shown).
In addition to the 25-kDa protein, both EMP-1 antisera detect a similar array of larger proteins in the intestine (Fig. 6B  and data not shown). The presence of these additional immunoreactive species varies from experiment to experiment and between tissues (data not shown). In general, the additional bands are most prominent in lysates containing higher amounts of EMP-1 protein. Since the two antisera are directed against independent regions of EMP-1 protein, these larger immunoreactive species are likely to represent aggregated molecules, a phenomenon frequently seen with highly hydrophobic proteins.
Although the level of EMP-1 protein observed in some tissues does not strictly correlate with EMP-1 mRNA expression, EMP-1 protein can only be found in tissues where EMP-1 mRNA expression is seen. No immunoreactive proteins are detected by either antiserum in lysates of the EMP-1 mRNA-negative liver (Fig. 6A).
Detection of Recombinant EMP-1 Protein by Immunofluorescence in Transiently Transfected COS Cells-To test the suitability of the polyclonal antisera to detect EMP-1 protein on frozen sections of rat tissue, we first assessed the specificity of the anti-loop 1 and anti-loop 2 antisera in COS cells transiently expressing EMP-1. COS cells were transfected by electroporation with a construct containing the EMP-1 cDNA under the transcriptional control of the cytomegalovirus promoter. Fortyeight hours later, cells were fixed, and antisera were used to detect EMP-1 expressing cells. After transfection, both antisera identified approximately 25% of the cells with strong immunoreactivity (Fig. 7). Neither antiserum nor preimmune sera recognized control COS cells transfected with the parental expression vector without EMP-1 insert (data not shown). Since detection with anti-loop 2 antiserum was more efficient than with the anti-loop 1 antiserum, the former antibody was predominantly used in subsequent studies. Fig.  8a shows a schematic representation of the topology of the rat gastric mucosa. The epithelial cells of the gastric pit are produced from stem cells in the isthmus/neck region of the gastric gland. These cells differentiate during their migration toward the gastric pit from which they are extruded (exfoliated) from the tip of the vilus (reviewed by Gordon and Hermiston (44)). Transverse sections across the gastric pit show epithelial cells that are organized in circles around the intestinal lumen (Fig.  8a).

EMP-1 Is Expressed in Epithelial Cells of Rat Intestine-
Ten-m frozen sections of corpus gastricum were stained with anti-loop 2 antiserum. Strong immunoreactivity was detected in the outer epithelial cells of the gastric mucosa from the tip of the vilus down toward the isthmus and neck of the gastric gland (Fig. 8, b and c). In transverse section, the EMP-1 immunoreactivity appears to be associated with the plasma membrane of epithelial cells in the gastric pits (Fig. 8e). The epithelial cells deeper in the gastric gland show little or no immunoreactivity, and specific labeling was not detectable in the base of the gastric gland or in the submucosal muscle layer (Fig. 8b). DISCUSSION We report the cloning and characterization of the epithelial membrane protein EMP-1, a hydrophobic polypeptide of 160 amino acid residues. Computer-aided analysis revealed that EMP-1 shows 40% amino acid identity to PMP22, a PNS myelin protein that is responsible for inherited peripheral neuropathies. EMP-1 and PMP22 display similar hydrophobicity profiles, suggesting that both proteins contain four membraneassociated, potentially membrane-spanning domains. Thus, we propose that EMP-1 and PMP22 are two members of a gene family that also includes the lens fiber cell protein MP20, one of the major protein components of the mammalian eye lens (33,40,41), as a distantly related third family member.
The high degree of identity at the amino acid level suggests that EMP-1 and PMP22 may serve similar functions. Close examination of the amino acid sequences of these proteins reveals that the hydrophobic regions, in particular the first two transmembrane domains, are highly conserved, suggesting that they are of particular functional importance. This hypothesis is further supported by the finding that the hydrophobic domains are the most strongly conserved regions between PMP22 species homologues. Interestingly, the amino acid residues in PMP22 that are sites of mutation in hereditary peripheral neuropathies are located within putative transmembrane domains, and the majority of these mutated amino acid residues are also conserved at the corresponding positions of EMP-1 and MP20.
The N-terminal signal peptides of both EMP-1 and PMP22 contain consensus sequences for signal peptidase cleavage. However, the signal sequence of PMP22 is not cleaved efficiently in myelinating Schwann cells as demonstrated by Nterminal sequencing of purified PMP22 protein (37). Since the N terminus of EMP-1 is also not cleaved when synthesized in the presence of CMM, we hypothesize that the EMP-1 signal peptide is not removed during biosynthesis in vivo. This situation is reminiscent of the structurally related connexin family of gap junction proteins, where a specific mechanism has been postulated to prevent aberrant N-terminal processing (31). Furthermore, MP20 does not contain a signal peptide cleavage consensus sequence, and N-terminal sequencing has shown it to be unmodified at its N terminus in vivo (42).
The most interesting conservation within the putative extracellular domains of EMP-1 and PMP22 concerns the consensus sequence for an N-linked glycosylation. This glycosylation site in PMP22 carries a modified carbohydrate chain containing the L2/HNK-1 epitope, a structure which has been implicated in cell-cell recognition and adhesion processes (for recent review see Schachner and Martini (1995)) (43). Although the presence and nature of carbohydrate moieties linked to EMP-1 remains to be determined, an N-linked glycosylation in the identical position of EMP-1 may be involved in cell recognition processes in the epithelium of the intestine.
EMP-1 and PMP22 are co-expressed in a wide range of tissues, and particularly high levels of transcripts for both proteins are found in the intestinal tract. The gastrointestinal tract is characterized by a continual and rapid renewal of its epithelial surface that continues throughout the animal's life. Pluripotent stem cells anchored in the isthmus/neck regions of the gastric gland give rise to progeny displaying increased proliferation and reduced potentiality, which progress to terminally differentiated mature cells (44). During this differentiation process, the cells are highly migratory, with prolifera-tion, migration, and differentiation all being tightly coupled. EMP-1 is found mainly in the proliferation and differentiation zones of the outer gastric gland as well as in the mature epithelial cells of the gastric pit region. In these cells, EMP-1 appears to be associated with the plasma membrane, with no clear distinction between the basal, apical, and lateral aspects.
PMP22 has been suggested to play a role in the control of cell proliferation. In support of this hypothesis, evidence has been presented that modulation of PMP22 levels in cultured Schwann cells has a pronounced influence on the cell cycle (28). In these experiments, overexpression of PMP22 increased the length of the G 1 phase, while reduced expression resulted in a decrease. Whether there is a similar effect of EMP-1 expression on the cell cycle of epithelial cells remains to be determined.
In this report, we confirm previous results that PMP22 is up-regulated in NIH 3T3 fibroblasts by growth arrest. In contrast, we show that EMP-1 mRNA levels are down-regulated under the same conditions. A short cDNA fragment corresponding to part of the 3Ј-untranslated EMP-1 mRNA has been isolated previously and, in agreement with our results, has been shown to be up-regulated in 3T3 fibroblasts 8 h after serum-induced entry into the G 1 phase (45). The same conditions lead to a decrease of the PMP22 mRNA level (20). In addition, we also observed that EMP-1 mRNA is up-regulated in the proliferating cells of the distal stump of the rat sciatic nerve 4 days after cut injury while PMP22 is down-regulated. Thus, PMP22 expression is induced by growth arrest in various experimental paradigms, while EMP-1 expression is decreased. This conspicuous inverse regulation of PMP22 and EMP-1 during the cell cycle lends further indirect support to a role of this protein family in the control of cell quiescence and proliferation. This hypothesis is particularly intriguing given the clinical interest in colon biology due to the prevalence of carcinomas in this region of the intestinal tract.
We have identified the major lens fiber protein MP20 as a distant member of the PMP22/EMP-1 gene family. MP20 has been described as a lens-specific membrane protein that colocalizes with connexin 46 in fiber cell junctions, suggesting a role in organizing the junctional plaques (46). MP20 has been shown to be absent from proliferating epithelial cells of the lens, with expression becoming prominent in differentiating as well as in mature lens fibre cells (33,46). Although the function of MP20 has not been elucidated, a role in signal transduction has been postulated, since MP20 is phosphorylated by cAMPdependent kinase in vitro and binds calmodulin in overlay assays (47,48).
With respect to the biological function of the PMP22/EMP-1/MP20 protein family, it is informative to discuss the genetics of PMP22. PMP22 has been intensively studied due to its association with the most common forms of hereditary peripheral neuropathies. Based on these studies and evidence provided by spontaneous and artificially generated PMP22-defective mice, it is established that PMP22 plays a crucial role in the development and maintenance of peripheral nerves. No obvious abnormalities outside of the nervous system are found in patients carrying heterozygous duplications, deletions, or point mutations affecting the PMP22 gene or in transgenic mice completely devoid of PMP22. The latter findings are difficult to reconcile with the widespread expression of PMP22 and its potential role in cell proliferation. However, the identification of the PMP22/EMP-1/MP20 gene family offers a plausible hypothesis for this apparent paradox. The function of PMP22 in non-neural tissues may be partially redundant, and compensatory mechanisms may be at work in mutant organisms. Correct PNS myelination appears to be the obvious exception to the rule, since the function of PMP22 in this tissue is critical. It is conceivable that the exquisite sensitivity of the PNS to PMP22 gene dosage, as inferred from human genetics and transgenic mice, does not allow for compensation. Alternatively, PMP22 is extremely highly expressed by myelinating Schwann cells, and the regulatory elements controlling potential compensatory genes may not be able to adjust expression accordingly in the mutant organisms. This hypothesis is also supported indirectly by the complex regulation of the PMP22 gene involving both a Schwann cell-specific and a ubiquitously active promoter (24). However, it cannot be excluded that PNSspecific functional peculiarities of the PMP22 protein are responsible for its crucial role in peripheral nerves.
These findings are reminiscent of the crystallin protein family (reviewed in Ref. 26). Crystallins are expressed in many tissues as enzymes or stress proteins. Through various evolutionary processes, they have subsequently been recruited as structural components of the lens. In some cases, high expression in the lens has been achieved using tissue-specific promoters or enhancers. Whether such mechanisms also apply to other known or yet to be found PMP22/EMP-1/MP20 family members remains to be seen.
In conclusion, PMP22 has been widely regarded as mainly a structural component of PNS myelin. Our description of EMP-1 and the concomitant identification of the PMP22/EMP-1/MP20 gene family, the expression pattern of these proteins in mainly epithelial tissues, and the observed differences in regulation of PMP22 and EMP-1 during the cell cycle support the concept that these proteins play multiple roles in cell biology. These functions may be related to both the switch from proliferation to differentiation as well as the maintenance of critical functions in the differentiated state.